MAX_WORDS=2900 was calibrated to the corpus median density and carried no
margin: at the densest observed 7.22 chars/word (~1.81 tokens/word) it permits
~5,240 tokens against the 5,000 it proxies for. 2770 holds the worst observed
density under the ceiling. The largest SKILL.md is 2,489 words, so the change
costs nothing today — 281 words of margin — and the header comment now argues
the new calibration rather than swapping the digits.
Both enforcement points move together, and a new test asserts they agree, since
a SKILL.md passing its own audit while the commit hook blocks it is the
disagreement this pair exists to prevent.
CONTEXT.md is deliberately left ungated: it is 2,816 words, and gating it would
block the build. Recorded here so the omission reads as a decision rather than
an oversight.
skill-audit's manual-fallback path listed only the line ceiling, so an agent
taking that path passed an oversized SKILL.md the hook then rejected. The word
ceiling is now named alongside it. agent-audit is deliberately unchanged: the
size hook scopes to SKILL.md only and agent-audit's validate.sh has no word
gate, so claiming it there would be false.
The Vale research doc still showed the MDX {/* vale off */} form under a
Markdown heading, contradicting CONTEXT.md and vale-run's troubleshooting
reference — that form suppresses nothing in plain .md. Fixed in both places it
appeared.
tests/run-tests.sh used mapfile (bash 4.0+) with unguarded array expansion,
though AGENTS.md tells contributors to run it and macOS ships bash 3.2. It now
collects via a while-read loop over process substitution and guards every
expansion. The newline-delimited find|sort pipeline is kept rather than -print0
with sort -z, whose BSD portability is the weaker link, and which matches
mapfile -t's previous behaviour exactly.
Refs: #85
ADR: 0013
The flattener's last-resort branch rewrote ASCII ' to U+2019, justified as the
one combination no YAML scalar can carry verbatim. That claim was false: a |-
literal block with a single indented content line carries ', ", \ and ": "
verbatim and keeps text.frontmatter.description matching — as the wrapper's own
docstring already said of literal blocks. The rewrite fired on 12 of 54
in-scope files, silently disabling every rule whose token contains an
apostrophe. Case 20 pinned only that the scope stayed alive, so it passed
either way.
The emission site now splits the emitted scalar on its first newline so a
carried-over trailing comment stays on the "description: |-" header line rather
than becoming part of the value, and pads by span_lines - 1 - newlines. The pad
stays non-negative because the branch is only reachable when the original span
is at least two lines. Verified across all 73 in-scope files: no line-count
changes, and exactly the 12 expected files take the new branch.
One reported position moves: an alert on a description that is itself flagged
shifts from the key line to the block's content line, both inside the original
span. YAML cannot put a literal block's content on the key's own line, so this
is unavoidable; no line at or after the end of any description span moves.
Also: --output no longer absolutises the built-in style names line, JSON and
CLI, which a same-named file or directory in cwd turned into a template path
(exit 2, E100 Runtime error). And case 19's empty-baseline guard no longer
lets five dependent comparisons print vacuous passes — while fixing it the
guard turned out to be unreachable, since under pipefail an alert-free report
aborted the script at the assignment.
Refs: #85
ADR: 0014
Three divergences between what the audit skills claim and what the hooks
enforce, each of which fails silently rather than loudly:
- `skill-size-check.sh` blocked at 2900 words while `validate.sh` checked only
the 500-line ceiling, so `/skill-audit` could report a skill ready to ship
that the commit hook then rejected. `validate.sh` now checks the same pair on
the same inclusive terms; the constants are duplicated with a comment naming
the other file, because a plugin skill's scripts cannot read outside the
plugin directory once installed to the cache.
- Both audit skills' Step 1 passed `--config assets/vale/.vale.ini`, which is
redundant (the wrapper self-locates its sibling config) and fragile: an agent
that resolves the script path against the skill directory but not the config
path gets E100, exit 2, which the surrounding fallback clause misreads as
"vale unavailable" and downgrades to full LLM judgment with no signal.
- The external-consumer test registered only the two Vale hooks, never the
third shipped hook, so a lost executable bit would have broken every consumer
while the local suite stayed green. Verified by mutation: `chmod 644` on the
copied script now turns three passes into two failures.
Also corrects the size hook's calibration comment, which claimed ~5.7-6.5
characters per word against a corpus whose measured median is 6.79 — the stated
upper bound sat below the median, so the "calibrated with margin" claim was
inverted for prose-dense files. MAX_WORDS is unchanged pending a decision; the
comment is now explicit that the gate holds under 5,000 tokens for typical
prose density, not for any file.
Refs: #85
Vale locates a frontmatter description by matching the parsed YAML value back
against the source text, so any scalar whose value is not spelled out verbatim
loses the `text.frontmatter.description` scope entirely. The wrapper only
flattened `>` folded scalars, so plain, double-quoted and single-quoted
multi-line descriptions silently reported zero alerts and exit 0 — a clean pass
indistinguishable from a real one, in a prefilter whose callers are instructed
not to re-derive its verdict by judgment.
Implementation notes:
- Classify the scalar kind after `^description:[ \t]*` and reuse one shared
continuation-line generator for every form; `|` literal blocks keep their
line breaks, stay verbatim-matchable, and are still left untouched.
- Emit the flattened value in whichever scalar form needs no escape at all
(plain, then single-quoted, then double-quoted), because any escape breaks
the verbatim match. The old blanket `'` -> U+2019 substitution silently made
apostrophe-bearing rule tokens unmatchable across 63% of the corpus; it now
survives only for the one combination no YAML scalar can carry verbatim.
- Terminate continuations at a line flush with the key, not only on a shallower
indent — a `description:` followed by a flush-left line previously swallowed
the rest of the frontmatter.
- Route vale's value-taking flags explicitly instead of inferring targets by
file existence, and absolutize relative `--output`/`--path` values the way
`--config` already was, since the run `cd`s into the scratch mirror.
- Fail loudly on a nonexistent path instead of inheriting bare vale's fallback
to stdin, which rendered a typo'd path as `0 errors ... in stdin`, exit 0 —
a form the callers' `0 files` NOT-RUN guard cannot match.
- Follow symlinks when walking a directory argument, matching bare vale.
Refs: #85
Under set -u, "${arr[@]}" on an empty array aborts on bash before 4.4,
which is what macOS ships as /bin/bash. Three expansion sites now use
${arr[@]+"${arr[@]}"} consistently.
The hazard is not currently reachable: verified on a bash 3.2.57 built
from source that all seven invocation shapes succeed against the
previous code, including zero args, flags-only and an empty directory.
vale_args is provably non-empty at every site because the default
--config branch always appends first. The guard is kept because that
invariant is non-local and untested, so an edit to the default-config
branch would reintroduce a macOS-only crash silently.
Test fidelity is deliberately mixed. Case 16 is static and is the only
one that fails against the previous code, since no bash 5 host can
reproduce the abort at runtime. Case 17 runs the emptiest invocations
under the oldest bash it can find and names that shell in its output
so it cannot overclaim. Case 18 guards against the tempting wrong fix
of dropping the quotes, which also silences the abort but word-splits
a path containing a space.
No other bash 4.x construct is present; swept for mapfile, declare -A,
case modification, negative indices, globstar, wait -n and namerefs.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MCQ648fLSFXPHGZdQ8gn58
pre-commit prefixes only entry[0] with the hook-repo clone path
(cmd = (prefix.path(cmd[0]), *cmd[1:])), so the --config argument in
.pre-commit-hooks.yaml resolved against the *consuming* repo's root
and hard-failed every external run with E100. Two of the three hooks
ADR-0014 promises were unusable.
vale-wrap.sh now self-locates its config from BASH_SOURCE when no
--config is supplied; an explicit --config still wins in all three
argv forms and stays cwd-relative. Both manifests drop the argument
and are kept byte-identical: the local repo: local config resolved
--config correctly only because the consuming repo *was* this repo,
and that divergence is why three review rounds missed the defect.
Also in the wrapper:
- replace GNU-only `realpath -m` with a portable abspath helper; -m is
load-bearing (dest does not exist yet), so BSD realpath aborted the
script under set -e on macOS
- walk directory arguments instead of passing them through unflattened,
which reported a clean 0-error run for files that fail when named
explicitly
- read/write with errors='surrogateescape' so one non-UTF-8 .md under a
directory argument cannot abort the hook
New test-vale-hooks-consumer.sh builds the hook repo from the working
tree and points a file:// consumer at it, covering the manifest as a
hook repo for the first time.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MCQ648fLSFXPHGZdQ8gn58
skill-audit/agent-audit's Step 1 resolved vale-wrap.sh/.vale.ini via
`git rev-parse --show-toplevel`, which returns whichever repo the skill
happens to run in. Inside ai-development that works; in any external repo
that installs kyberforge@holocron as a plugin, it resolves to that repo's
own root, which has no .vale.ini — the prefilter silently fell back to
full LLM judgment. ADR-0013 named this as a deliberately deferred gap.
Vale's config/styles/wrapper now ship inside the plugin itself: a
canonical copy in agent-audit/assets/vale/ (Kyberforge + KyberforgeCopilot,
the superset agent-audit needs) and a smaller duplicate in
skill-audit/assets/vale/ (Kyberforge only) — per the no-cross-skill-path
rule already established for plugin cache-installs. Both skills resolve
these relative to their own directory, same as scripts/validate.sh
already does.
A new root .pre-commit-hooks.yaml exposes both copies plus
skill-size-check so any external repo can enforce the same rules via
`repo: <this-repo-url>, rev: <tag>` in its own pre-commit config,
independent of Claude Code entirely — the same mechanism covers CI. This
repo's own pre-commit hook now consumes the identical plugin-bundled
copies via repo: local (not a third root copy, and not a pinned
self-reference, which would lint working-tree edits against the last
tagged release instead of the change being made). Split into
vale-audit-prefilter-skill/-agent hooks after confirming, by diffing the
full corpus against both old and new config before deleting the old
files, that one combined hook pointed at only one copy silently 0-file-
skips the other file type.
scripts/check-vale-style-sync.sh guards the two copies against drift,
wired at pre-push alongside check-manifests.
ADR: 0014
Round-3 review of PR #85 found the "enforcing" pre-commit hook enforced
nothing. Vale's exit code keys on error-level alerts alone: five of the
six rules were level: warning, so they exited 0, and pre-commit hides
output from a passing hook — the alerts were invisible and blocked
nothing. ADR-0013 rejected a report-only trial tier and then shipped one
by accident.
Flatten every rule to level: error. Vale's own exit code is then correct,
so the hook entry drops to a bare vale-wrap.sh call and the graded
error->FAIL / warning->SUGGESTION mapping disappears from both audit
skills: every alert is a FAIL, in the gate and the audit alike. No
ignorable tier, matching shellcheck, the test suite and
conventional-pre-commit.
Delete Kyberforge.VagueQualifier. Measured against the 41 skill/agent
files as they stood before the rule ever ran: 2 hits. One marginal
("very different" -> "fundamentally different"), one an unfixable false
positive — caveman/SKILL.md quotes "of course" as an example of filler,
a mention not a use — which forced the only Vale suppression comments in
the repo. Those four lines go with it; two of them were dead anyway,
suppressing a frontmatter-scoped rule on a body line. Held-out prose (273
files) fired 15 times, 9 inside out-of-scope research examples and the
rest one word in two idioms in a single doc. SentenceOpenerThereIs
survives: 22 held-out hits, both in-corpus hits clean rewrites, zero
suppressions.
Widen .vale.ini's globs to [**/SKILL.md], [**/agents/*.md] and
[**/*.agent.md]. The plugins/*/-prefixed globs scoped nothing — Vale's *
crosses /, so they already matched docs/research/examples/**/agents/*.md
and assets/templates/SKILL.md, the two paths CONTEXT.md claimed they
excluded. Scoping is and was the hook's files: regex. The old globs also
hid a silent false negative: a skill outside plugins/ matched no section,
so Vale reported 0 files and exited 0, which both audits read as clean.
They now treat a 0-file run as NOT RUN and fall back to full judgment.
Also:
- vale-wrap.sh resolves relative --config values and file arguments
against the caller's cwd, as vale does, instead of the repo root, which
hard-errored from a subdirectory and silently skipped flattening for
file args that did not resolve from the root. Absolute paths inside the
cwd are relativized so reports cite resolvable paths, not scratch ones.
- vale-run's exit-code model was documented backwards ("exits non-zero
whenever it finds an alert at or above MinAlertLevel") and would have
led anyone following it to build a gate that passes everything. Its
Markdown suppression syntax was MDX-only and does not suppress in .md;
corrected in the skill and its troubleshooting reference, with
backtick/fence exemption documented as the first resort.
- skill-size-check.sh fails only above 500 lines, agreeing with
skill-audit's validate.sh <= 500 pass.
- ADR-0013 and CONTEXT.md amended to match, recording why graded
severities cannot gate.
Verified: 9 test scripts / 15 vale-wrap cases pass; vale-audit-prefilter,
skill-size-check and shellcheck pass --all-files; check-manifests and
claude plugin validate --strict clean. New tests fail against the old
script (3 of them) and pass against the new one.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MCQ648fLSFXPHGZdQ8gn58
Addresses PR #85's outstanding review items after grilling the open
questions against ADR-0013/CONTEXT.md/ADR-0010:
Blocking fixes:
- vale-wrap.sh: replace json.dumps() escaping (which silently defeated
Vale's frontmatter scope on any description containing a quote,
backslash, or non-ASCII char — ~58% of the corpus) with a single-quoted
YAML scalar, substituting a Unicode right single quote for embedded
apostrophes rather than '' doubling (Vale's frontmatter scanner isn't a
full YAML parser and silently truncates on '' too).
- vale-wrap.sh: fix a blank-line-inside-a-folded-description truncation
bug via indentation-based, blank-line-tolerant body capture; narrow
flattening to `>`-style scalars only (`|` already works unflattened).
- skill-audit/agent-audit Step 1: make the vale-wrap.sh invocation
cwd-independent via git rev-parse --show-toplevel, fixing a bug where
no single cwd satisfied all three Step 1 commands.
- styles/Kyberforge/VagueQualifier.yml: prune 17 tokens verified
false-positive-dominated on this repo's own voice via a real corpus
sweep (obvious, clearly, usually, several, simple, easy, completely,
simply, tiny, etc.), keep 13 with real or unattested noise. Revert the
28 prose "fixes" those tokens drove across 14 skill files back to their
original, correct wording, including a functional regression to
caveman/SKILL.md's own filler-word list (a mention, not a use) — now
guarded with vale-off comments against recurrence.
Gaps:
- --minAlertLevel=warning on the pre-commit hook and Step 1 invocation
so warning-level rules actually surface, without collapsing the
FAIL/SUGGESTION severity mapping skill-audit/agent-audit rely on.
- vale-wrap.sh: fix --config=<path> equals-form, absolute-path silent
no-op, and a zero-file-argument stdin hang.
- Route vale-run and lint-runner through a documented wrapper script
when a target repo has one, instead of unconditionally recommending
bare `vale`.
- Wire Kyberforge.VagueQualifier/SentenceOpenerThereIs into skill-audit/
agent-audit's dimension-mapping prose (Body discipline).
- Add plugins/lint/sources.md provenance for lint-runner (ADR-0010).
- Sync both marketplace.json lint-entry descriptions with plugin.json.
- Retune skill-size-check.sh's MAX_WORDS 5000->2900 (measured ~1.6-1.7
tokens/word on this repo's corpus, the old value gated at ~8,500
tokens against a stated 5,000 ceiling); fix the >/>= line-count
boundary and wc -l undercount on files with no trailing newline.
- Document the vale binary as a Setup prerequisite in AGENTS.md.
- Fix SentenceOpenerThereIs's dead regex alternative and add a real
sentence-start anchor/scope.
- Fix a stale docs/research/docs/vale/ index pointer in kyberforge's
docs README (moved to plugins/lint/ in e1a5403).
- Rewrite ADR-0013's Consequences section past-tense to describe what
actually landed, and record the styles-portability limitation
(repo-root placement stays intentional; deferred to a separate
session per this PR's review).
Test coverage: 9 new vale-wrap.sh fixtures (quotes, backslash/unicode,
blank-line paragraphs, --config= form, zero-arg/absolute-path handling,
literal-block no-regression) and boundary-pair tests for
skill-size-check.sh's line/word ceilings.
bash tests/run-tests.sh: 9 scripts + 125 bats assertions, all passing.
scripts/check-manifests.sh and claude plugin validate --strict: clean.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MCQ648fLSFXPHGZdQ8gn58
Deferred item from PR #85 review. Per ADR-0013: cherry-picks two low-noise
rules from trialing write-good/alex against the real corpus (VagueQualifier,
SentenceOpenerThereIs) into styles/Kyberforge rather than adopting either
package wholesale (both are tuned for blog prose and were noisy on this
repo's terse, imperative instruction files - see the ADR's rejected-rule
list). Adds a new skill-size-check pre-commit hook enforcing agentskills.io's
500-line/5,000-token SKILL.md ceiling, currently unenforced. Fixes the 28
resulting violations across 20 existing SKILL.md/agent files so the
enforcing pre-commit hook lands clean.
governance.md/CONTROLS.md were evaluated and excluded as rule sources -
they're org/CI-infrastructure controls, not prose patterns Vale can express.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QUDczvw1H3eEeMD29Q9Lbi
Vale's text.frontmatter.description scope silently stops matching once
the description is a YAML block scalar spanning 2+ physical lines —
the style used by most skills/agents in this repo. scripts/vale-wrap.sh
flattens the description to one line in a scratch copy (preserving the
repo-relative path and total line count) before invoking real vale, and
both audit skills plus the pre-commit hook now call it instead of vale
directly. Also tightens the pre-commit hook's file glob to single path
segments so it can't cross into docs/research examples or asset
templates the way the audit skills' scoped invocations already avoid.
Addresses PR #85 review feedback.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FxG5T8EJDgkABXxuneuFfn
Adds repo-root .vale.ini plus a custom Kyberforge style (description-opener,
vague-wording, and generic reference-pointer padding rules) and a
KyberforgeCopilot style scoped to .agent.md files (Use proactively check).
skill-audit and agent-audit Step 1 now run vale against the specific file(s)
being audited and defer the corresponding Description/Patterns/Body checks
to its output instead of re-deriving them by LLM judgment, per the split
proposed in issue #84.
Closes#84
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FxG5T8EJDgkABXxuneuFfn
The Prerequisites checklist mixed items to confirm (preconditions) with an
action to perform (capturing git log), so the following "stop and ask if
missing" gate didn't logically apply to the git-log step. The improve flow
also had no reminder to update source_keys/sources.md when an edit touches
research-sourced content, unlike the create flow's explicit step for it.
Refs #69
Independent skill-audit found that agent-author's closing checklists never
verified template <!-- --> comments were stripped from frontmatter (produces
invalid YAML if left in), the Copilot field-exclusion checklist omitted two
fields present in the authoritative list, and the improve flow had no
agent-audit availability check unlike the create flow.
Prior sessions had authoring subagents report completion after only
staging changes (git diff --stat showing output, but no git commit).
agent-author's create and improve flows now require capturing
git log --oneline -1 before and after the authoring pass and asserting
the hash actually changed via a real commit, matching the fix already
applied to skill-author.
Refs #69
Independent /skill-audit recheck flagged the git-log-capture instructions
as discoverable only at close-out (Step 6/Step 5), long after the step
where the hash should actually be snapshotted. Adds the capture checklist
item to Prerequisites (create flow) and Step 1 (improve flow) instead of
leaving it as a retrospective-only note. Also drops a sentence in the
improve flow's Step 4 that duplicated the preceding one on editing
scripts/reference files directly.
Refs #69
Prior sessions had authoring subagents report completion after only
staging changes (git diff --stat showing output, but no git commit),
and one run self-spawned its own audit/recheck subagent instead of
leaving that to forge's outer loop, losing an uncommitted draft when
the stray subagent's worktree was torn down.
Refs #69, #71
skill-audit/agent-audit now proactively trigger after a skill/agent
file is hand-edited outside skill-author/agent-author, not just on
explicit request — closing a gap from this session where a fork's
direct edits to agent-author/agent-audit shipped without their own
inline audit until forge was invoked to check afterward.
Also: plugin-author gains a gotcha on claude plugin validate --strict
auto-discovering every .md under agents/ regardless of manifest
declarations (ADR-0010); agent-audit's dimension count and
validate-provenance.sh's --help now match actual behavior.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Independent skill-audit clean-check surfaced that the dimension-mapping
list and manual-fallback checklist both skipped the counterpart-missing
case, which validate.sh already treats as a hard FAIL.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
`claude plugin validate --strict` auto-discovers every .md under a
plugin's agents/ directory as an agent requiring frontmatter, so the
provenance file there always needs fake agent frontmatter to pass
validation. Confirmed empirically that an explicit `agents` manifest
array can't suppress this discovery. Move the file to <plugin-root>/
sources.md instead, and update agent-author/agent-audit accordingly.
Adds ADR-0010, partially superseding ADR-0005's `agents/sources.md`
convention. Fixes#63.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
## Why
The git plugin only covered a partial slice of common git workflows.
This adds the remaining skill set (branches, commits, history, remotes,
submodules, workflow, worktrees) plus a git-orchestrate agent so the
plugin can handle end-to-end git automation instead of a handful of
commands.
## Implementation Notes
Each new skill was validated against its research docs and org
conventions after initial authoring, which surfaced hallucinated
version pins, factual errors, and completeness gaps that were
corrected in the same pass rather than left for follow-up.
## Impact
Bumps the git plugin to 1.3.0.
Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>
Step 4 ensures that after creating/updating artifacts in a plugin, forge invokes plugin-author (clean-context subagent) to bump the plugin version if not already done by other skills. Prevents missed version updates when artifacts are added to plugins.
Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
Grills intent, classifies target artifact type (skill/agent/plugin/marketplace
entry) against a fully descriptive table, then routes to the matching author
skill via fork subagent (falling back to inline when fork is unavailable or
the flow needs live interaction). Adds an independent clean-context audit
recheck after each skill/agent route, looping author-then-audit until the
recheck comes back clean, since the author skill's own inline audit shares
context with the work it verifies. Updates CONTEXT.md's Skill composition
entry to describe this recheck loop and adds forge's provenance chain
(references/sources.md).
Refs Defame1297/holocron#61
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
- agent-author: convert template comments from YAML (#) to HTML (<!-- -->)
- Easier to spot and distinguish from functional comments
- Add explicit "Delete template comments before shipping" reminders
- Update SKILL.md Steps 2-3 with removal instruction
- agent-audit: add comment-discipline check
- Flag excessive frontmatter documentation comments as padding
- Mirrors skill-audit's body-discipline principle
- Update coverage line to include comment-discipline dimension
This ensures agents follow the same comment-cleanup discipline as skills,
preventing template documentation from shipping with agent definitions.
Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
## Why
Two issues were found in `validate-provenance.sh` by `/skill-audit`:
1. `find_plugin_root` only walked up looking for bare `plugin.json`, missing
the `.claude-plugin/plugin.json` layout used by kyberforge plugins — matching
the logic already present in `validate.sh`.
2. Check numbers in `--help` and inline comments had a gap (0,1,2,4,5,6) from
a previously removed check, making the numbering confusing to readers.
## Implementation Notes
Check numbers renumbered sequentially 0–5 in both the `--help` output block
and the inline `# --- Check N:` comments. The `find_plugin_root` condition now
mirrors the `detect_scope` function in `validate.sh`.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0147vXtL5sP6vorDdqXGJJU9
Corrects three FAIL findings from the skill-audit run:
- Scope detection table in SKILL.md and references/scripts.md now lists all
four plugin-marker variants the script actually checks (plugin.json,
.claude-plugin/plugin.json, .plugin/plugin.json, .github/plugin/plugin.json)
- assets/README.md copilot.agent.md description was wrong about field set;
replaced with accurate CLI-format description noting excluded cloud/IDE fields
and the Copilot tool aliases actually used.
Also applies the SUGGESTION: moves the conditional reference
(`If the destination is a plugin directory, read references/deployment-modes.md`)
out of the ## Gotchas section body and into ## Route as a standalone line,
immediately after ## Gotchas closes.
INFO findings (source_keys frontmatter) were already present in both
references/README.md and references/scripts.md — no change needed.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0147vXtL5sP6vorDdqXGJJU9
## Why
The description sentence "Handles both the full create flow (scaffold → fill →
validate) and the improve flow (signals → root cause → edit → audit)." describes
internal mechanics rather than user intent. The surrounding trigger phrases already
cover both create and improve intents, so this sentence adds no triggering value
and violates the description principle of focusing on what the user is trying to
achieve.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0147vXtL5sP6vorDdqXGJJU9
Remove phantom cross-file name-match claim from SKILL.md manual fallback
list (validate.sh has no such check). Fix broken bats stem-mismatch test
to overwrite the Copilot file instead of the CC file, exercising the
actual Copilot stem check.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0147vXtL5sP6vorDdqXGJJU9
- Fix validate.sh: remove incorrect name==stem check for CC files (CC docs say filename need not match name field); keep check for Copilot CLI only
- Fix validate.sh: plugin scope detection now checks both plugin.json and .claude-plugin/plugin.json
- Fix validate.sh: Copilot cloud/IDE agents (.github/copilot/agents/) have name as optional; path-based guard added
- Add validate.sh checks: Copilot body length >30,000 chars (SUGGESTION), Copilot-only fields in CC files (FAIL), subagent-unavailable tools in tools field (SUGGESTION)
- Add references/description-quality.md as conditional escape hatch for borderline description findings
- SKILL.md: name five audit dimensions in description; sharpen indirect-trigger phrasing
- SKILL.md: label pair-mandate as kyberforge project convention, not platform requirement
- SKILL.md: scope redundant name-match/body-empty checks to manual fallback only
- SKILL.md: add conditional reference to description-quality.md; update provider-safety description for new check categories; fix plugin scope gotcha to mention .claude-plugin/plugin.json
- SKILL.md: add INFO tier to result block template
- Add source_keys frontmatter to references/README.md; update sources.md to add description-quality.md to contributing files
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0147vXtL5sP6vorDdqXGJJU9
- references/sources.md: add YAML frontmatter with source_keys to fix provenance chain break
- SKILL.md description: make create/improve duality explicit ("Handles both the full create flow ... and the improve flow ...")
- SKILL.md Step 2: add inline metadata.source_keys instruction — fill early, not deferred to Step 5
- SKILL.md Step 6 (create) / Step 5 (improve): add pre-audit manual checklists and version bump conventions (minor for create, patch for improve)
- README.md: correct false claim that the skill bumps plugin manifests; it bumps metadata.version only
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0147vXtL5sP6vorDdqXGJJU9
## Why
agent-author produces agents/sources.md at plugin scope to record which
research sources informed which agent files. agent-audit had no way to
validate this chain, leaving stale or missing provenance undetected.
## Implementation Notes
Validation is per-pair (the given agent file + its counterpart) rather
than plugin-wide, keeping the scope consistent with validate.sh. The
script exits 0 silently for non-plugin-scope agents.
source_keys is top-level in both CC .md and Copilot .agent.md files
(not under metadata:) to avoid conflict with Copilot's own metadata
field semantics. Checks 0, 1, 2, 4, 5, 6 mirror the skill provenance
set; upstream research-doc cross-reference checks (7, 8) are deferred.
agent-author Steps 2, 3, and 4 updated to formally specify the
agents/sources.md format and instruct authors to add source_keys to
both files when research sources are in context.
Refs: #60
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0147vXtL5sP6vorDdqXGJJU9
## Why
The last acceptance criterion from #11: agent-author's close step should
reference agent-audit so authors are prompted to validate the pair before
shipping, matching the pattern skill-author uses with skill-audit.
Refs: #11
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0147vXtL5sP6vorDdqXGJJU9
## Why
agent-author produces paired agent definition files (Claude Code .md +
Copilot .agent.md) but had no companion audit skill to validate them.
agent-audit fills that gap, giving the same structured PASS/FAIL report
that skill-audit provides for SKILL.md files.
## Implementation Notes
- validate.sh uses scope detection (walk up for plugin.json / .git) to
locate the counterpart file and determine whether plugin-silently-ignored
fields (hooks, mcpServers, permissionMode) should be flagged
- CC-only and silently-ignored field lists are read from
references/field-inventory.md at runtime rather than hardcoded —
provenance back to the research corpus; see ADR-0019
- Single-file invocation (pass either file, counterpart derived) chosen
over directory or name+root — see ADR-0018
- 12 bats tests cover provider detection, scope detection, all FAIL paths,
and clean-pair pass
## Impact
- kyberforge bumped to v1.1.2
- agent-author close step should be updated to reference agent-audit (#11)
- Provenance/sources chain check deferred to #60
ADR: docs/adr/0018-agent-audit-single-file-invocation.md
ADR: docs/adr/0019-agent-audit-field-inventory-reference.md
Refs: #11
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0147vXtL5sP6vorDdqXGJJU9
## Why
The UPDATE flow only bumped version when the version field itself was the
target. Metadata changes (description, keywords, author) went out without a
version bump, making them invisible to consumers with a cached copy.
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
## Why
Same gap as skill-author: no instruction to bump the plugin version after
creating or modifying an agent inside a plugin.
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
## Why
README "What it does" didn't reflect the plugin version bump step added
to both create and improve flows.
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
## Why
The skill had no instruction to bump the catalog version after ADD, REMOVE, or UPDATE operations. Clients cache the marketplace catalog and use the version field to detect changes — without a bump, the new state is invisible until a forced refresh.
## Implementation Notes
Added a Gotcha explaining the rule and semver convention (ADD/REMOVE → minor, UPDATE → patch). Added a dedicated "Bump catalog version" step to each mutating flow. Updated README to match.
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Both skills exist on disk but were missing from the plugin README and the
skills/ directory README. Adding them so the listings stay consistent with
the actual skill set.
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
## Why
Several small quality issues accumulated across plugin-author and
marketplace-author: the plugin-author description was ambiguous about
scope, the ADD flow asked an unnecessary clarifying question when local
path is the obvious default, VALIDATE had redundant wording already
captured inline, the keywords field was missing from the validation
checklist, and tests/README.md was a placeholder with nothing to test.
## Implementation Notes
- plugin-author: tightened description scope to "inside those
directories"; added `keywords` identical-in-both-manifests check to
VALIDATE checklist.
- marketplace-author: moved reserved-prefix constraint from gotchas
into the ADD checklist item; replaced the four-option prompt with a
default+escape-hatch ("I'll treat this as local path — is that right?");
folded the validate-from-root note inline and removed trailing padding;
deleted tests/README.md — no scripts/ directory exists, so the
placeholder was noise.
- README.md: removed the tests/README.md row from the file table.
## Why
The manifest-fields tables in both skills used imprecise labels ("CC-only",
"Copilot-only") that conflated two distinct reasons a field appears in only
one manifest: platform constraint (the other tool does not support the field
at all) versus repo convention (both tools support it, but the scaffold places
it in one manifest by design). This caused agents to treat convention
boundaries as hard platform constraints, producing unnecessary errors when
updating manifests for dual-tool repos.
Provenance was also incomplete: sources.md files were missing entries for
sources that had been consulted and were already contributing to SKILL.md
and manifest-fields.md content, making the evidence chain unverifiable.
## Implementation Notes
Field classification now uses three explicit categories — shared, platform
(one tool does not support the field), and convention (both tools support it;
scaffold places it in one manifest by design). The distinction matters because
convention fields may legitimately appear in the other manifest when there is
a deliberate reason; platform fields may not.
New gotchas added to plugin-author: agent files silently ignore hooks,
mcpServers, and permissionMode frontmatter; claude plugin tag --push requires
a clean working tree; --dry-run preview before tagging; --strict flag on
validate. New gotchas in marketplace-author: metadata object as Copilot CLI
canonical location for top-level fields; strict: false for dual-tool plugins;
sha takes precedence over ref for pinning; --strict flag on validate.
tests/ removed from plugin-author because new-plugin.sh has no branching
logic warranting a bats suite at this stage.
## Impact
Skill prompt changes only — no runtime code affected. Agents using these
skills will now correctly distinguish convention from constraint when deciding
which manifest to update for a given field.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
## Why
Plugin and marketplace management had no governed authoring path. Creating or
updating a plugin required knowing the dual-manifest convention, version parity
rules, and directory skeleton by memory — nothing enforced consistency or guided
the process.
`/plugin-author` closes that gap by owning the full plugin scaffold lifecycle:
create, update, rename, and release. `/marketplace-author` handles the
marketplace-facing side: register, deregister, and update plugin entries in
`marketplace.json`.
ADR-0016 codifies the version parity convention (identical `version` in both
`plugin.json` and `.claude-plugin/plugin.json`) that `/plugin-author` now
enforces. The two plugin.json files in this repo are backfilled to comply
(keys also sorted to pass the pretty-format-json hook). CONTEXT.md gains
glossary entries for "plugin scaffold" and "version parity" so future agents
have shared vocabulary for these concepts.
## Implementation Notes
`/plugin-author` ships a `scripts/new-plugin.sh` scaffold script that generates
the directory skeleton and both manifests in one shot; the skill calls the script
rather than generating files ad hoc so the scaffold is reviewable and repeatable.
Version parity is an invariant, not a suggestion — the skill will fail loudly
on create/update if the two versions would diverge.
ADR: docs/adr/0016-plugin-version-parity.md
## Why
Pre-commit config management was entirely manual — no skill existed to
help create, modify, or validate `.pre-commit-config.yaml`, or to run,
install, and maintain the pre-commit setup. These two skills close that
gap with clear scope separation: authoring vs. execution.
## Implementation Notes
- `pc-author` owns `.pre-commit-config.yaml` only (no hook publishing,
no install). Runs `pre-commit validate-config` after every write.
Shallow file-extension scan drives proactive hook recommendations;
rev staleness is flagged against `references/hooks-by-language.md`
rather than hardcoded versions. Remove path reverts on failure.
- `pc-run` owns install, run, autoupdate, gc, and clean. Defaults to
`--all-files`. Install warns about existing `.git/hooks/` files being
overwritten by `-f`. Clean requires HITL confirmation. Failure
interpretation delegates to `references/failure-patterns.md`.
- Provenance wired to `plugins/kyberforge/docs/research/docs/pre-commit/`.
- Both skills resolve via the existing `"skills/"` glob in `plugin.json`.
## Impact
Two new slash commands available after `claude plugin install kyberforge@holocron`:
`/pc-author` and `/pc-run`.
Refs: #12
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01P87CiC58Ru2PPWYTeXtjHT
## Why
skill-author explicitly excludes agent definition files ("Do not use to author
agent definition files"). No factory skill existed to create or improve the
.md / .agent.md files that define Claude Code subagents and Copilot CLI agents
in a plugin, project, or user scope. This fills that gap.
## Implementation Notes
Single-root scaffold convention: new-agent.sh <name> <root> derives both
provider file paths from the root by convention — plugin scope (plugin.json
present) writes both files into <root>/agents/; non-plugin scope writes
.claude/agents/<name>.md and .github/agents/<name>.agent.md. This keeps
input minimal while always generating both provider files. See ADR-0015.
Routing is file-level (not directory-level like skill-author): neither file
exists → create flow; at least one exists → improve flow; scaffold is a
file-by-file no-op so retries are safe.
No companion agent-audit skill — inline validation in the close step covers
the simpler agent field contract. agent-audit is tracked as a follow-on.
## Impact
Closes the skill-author gap for agent definitions. Follow-ons tracked in
Gitea #11: agent-audit skill and --copilot-dest override flag for non-standard
Copilot project paths.
---
ADR: docs/adr/0015-agent-author-dual-provider-scaffold.md
Refs: #10
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
skill-audit description understated its coverage by omitting three audit
dimensions (patterns, scripts, provenance); README carried the same stale
list. skill-author had an unconditional reference trigger that should be
conditional on whether a script is being added.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Remove implementation-focused sentence from description (was not user-intent
language per agentskills.io spec)
- Move source_keys template comment under metadata: block to match Step 5's
instruction; contradicted agents scaffolding before reading Step 5
- Add Step 6 to new-skill.sh next-steps (populate references/sources.md);
renumber validate step to 7 — script was missing the sources step entirely
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
references/sources.md Research doc: fields are development-only provenance
pointers, not runtime references — they intentionally target paths outside the
skill directory and are expected to be non-resolvable after plugin install.
validate-provenance.sh degrades gracefully when they don't resolve.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
