Defame1297
ede3f06689
Diffing each retrofitted SKILL.md against its replacement references/ files found rules that existed on main and now existed nowhere — relocated in intent, deleted in fact. A trim that loses a rule is not progressive disclosure, it is data loss with a smaller word count. Three had no survivor. The least-privilege guidance for `tools` kept its mechanics and lost the "restrict to what the agent needs" half, so the remaining text read as encouragement to omit the field. The improve flow lost its regression check, so nothing compared the closing audit against the pre-edit state and a PASS quietly becoming a SUGGESTION went unnoticed — restored on both halves of the author pair, since agent-author had dropped its equivalent too. And agent bodies lost "would the agent get this wrong without it?", which mattered more than it looks: ADR-0020 deliberately sets no body word gate for agents, three of the four already sit between 933 and 1,199 words, and the delegation check only fires on procedure a skill already owns. That heuristic was the only brake left. Two more were reachable only from the wrong scope. agent-author tells the reader to load only the file for the resolved scope, but the mcp__ glob syntax for disallowedTools and the five tools no subagent ever receives had both landed in project-user-scope.md. disallowedTools is the ONLY permitted fence at plugin/APM scope, so the scope that needs the syntax most could not reach it, and a plugin-scope run could write a body telling the agent to ask the user a question. Two documents were actively wrong rather than merely thin. agent-audit told auditors that validate.sh resolves boundary targets for skills only; it runs at both scopes, so the auditor was hand-resolving what the script had already decided and could contradict it. And skill-audit routed to its script-troubleshooting reference whenever validate.sh "fails" — but it exits 1 on ordinary content FAILs, the normal outcome for the whole #99 population, so 1,302 words loaded on nearly every audit. A context-budget regression inside the skill that enforces the context budget. Finally, two illustrations taught the shape the gate ERRORs on, unfenced, while an adjacent rubric called it a hard ERROR. LESSONS.md records the reference-chain depth rule flipping from "one level deep" to "two hops, never three". ADR-0020 is silent on it and the reversal rode entirely on the diff; the looser rule is what mandatory dispatch requires. Refs: #99 ADR: 0020 Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015W3iwF9ncfRZddGBxsMCYi
kyberforge
Skills and agents for creating, maintaining, and managing a Claude Code / Copilot CLI plugin marketplace.
Install
Claude Code:
claude plugin marketplace add <owner>/<repo>
claude plugin install kyberforge@<marketplace-name>
GitHub Copilot CLI:
copilot plugin marketplace add <owner>/<repo>
copilot plugin install kyberforge
Local (development):
# Claude Code
claude --plugin-dir ./plugins/kyberforge
# GitHub Copilot CLI
copilot plugin install ./plugins/kyberforge
Contents
Authoring source lives in .apm/. The skills/, agents/, and hooks/hooks.json paths below are the generated mirror that plugin hosts actually scan — produced by scripts/sync-plugin-content.sh, never hand-edited (ADR-0017).
| Component | Path | Description |
|---|---|---|
| Skills | .apm/skills/ → skills/ |
Slash commands available after install |
| Agents | .apm/agents/*.agent.md → agents/ |
Role-based agents; one vendor-neutral .agent.md per agent, copied verbatim to both targets (ADR-0016) |
| Hooks | .apm/hooks/ → hooks/hooks.json |
Event-triggered automation — read by Claude Code only, see below |
| MCP servers | .mcp.json |
Model Context Protocol server definitions (hand-authored at the plugin root) |
Hooks are Claude Code-only in practice. Claude Code convention-scans hooks/hooks.json at the plugin root, so that file is live. Copilot CLI has no default hooks path — agents and skills default to agents/ and skills/, but hooks defaults to nothing (docs/research/docs/github-copilot-plugins/configuration.md:47), so Copilot reads hooks only via an explicit "hooks": "hooks.json" pointer in plugin.json (docs/research/docs/github-copilot-plugins/examples.md:49). apm's manifest compiler strips pointer fields from every manifest it generates, so neither .claude-plugin/plugin.json nor .github/plugin/plugin.json carries one, and re-injecting it is the option ADR-0017 explicitly rejected. Copilot therefore loads no hooks from this plugin — and did not load them from the old root-level hooks.json either. The Copilot half of this row is aspirational, not current behaviour.
Skills
| Skill | Description |
|---|---|
forge |
Grill an unclassified "I want to add something" request, decide whether it's a skill, agent, plugin, or marketplace entry, then route to the matching author skill |
skill-author |
Create or improve a skill from scratch, audit findings, or inline feedback |
skill-audit |
Audit a skill directory against the agentskills.io spec and produce a findings report |
agent-author |
Author an agent definition file |
agent-audit |
Audit an agent definition across structure, provider safety, description and body quality, and provenance; produces a findings report |
apm-install |
Install or upgrade the apm CLI and set up the agent runtimes it drives (Copilot CLI, Codex, Gemini, generic llm) |
apm-workflow |
Author apm.yml, scaffold an apm package/marketplace, install dependencies, and compile/pack/publish/audit apm content |
Pre-commit tooling (pc-author, pc-run) lives in the git plugin, not here.
Author
Defame1297