Summary
agents/README.md is placed inside the agents/ directory, which causes a harness warning in tools that scan agents/*.md and parse each file as a custom agent definition (e.g. GitHub Copilot CLI).
What happens
GitHub Copilot CLI scans every *.md file in a plugin's agents/ directory and attempts to parse it as a custom agent (which requires YAML frontmatter with at least a description). On agent-skills, the four real agents parse fine:
agents/code-reviewer.md ✅
agents/security-auditor.md ✅
agents/test-engineer.md ✅
agents/web-performance-auditor.md ✅
But agents/README.md starts with # Agent Personas and has no YAML frontmatter, so it fails to parse and emits a warning on every session start:
[WARNING] .../installed-plugins/addy-agent-skills/agent-skills/agents/README.md:
custom agent markdown frontmatter is malformed: missing or malformed YAML frontmatter
Impact
Cosmetic only — the four real agents load fine and no skill is affected. But the warning fires on every startup and can be misread by users as "an agent/skill failed to load."
Suggested fix (any one of these)
- Move the docs to
agents/PERSONAS.md won't help (same dir). Instead move it out of the scanned folder, e.g. docs/agents.md, and link to it from the top-level README.md.
- Or add minimal YAML frontmatter to
agents/README.md so it parses as a (harmless/no-op) entry — not recommended, as it would surface a bogus "agent."
- Document agents in the repo-root
README.md instead of shipping a README inside agents/.
Option 1 (relocate to docs/) is cleanest and keeps agents/ containing only real agent definitions.
Environment
- GitHub Copilot CLI 1.0.61 (macOS)
agent-skills plugin v1.0.0
Summary
agents/README.mdis placed inside theagents/directory, which causes a harness warning in tools that scanagents/*.mdand parse each file as a custom agent definition (e.g. GitHub Copilot CLI).What happens
GitHub Copilot CLI scans every
*.mdfile in a plugin'sagents/directory and attempts to parse it as a custom agent (which requires YAML frontmatter with at least adescription). Onagent-skills, the four real agents parse fine:agents/code-reviewer.md✅agents/security-auditor.md✅agents/test-engineer.md✅agents/web-performance-auditor.md✅But
agents/README.mdstarts with# Agent Personasand has no YAML frontmatter, so it fails to parse and emits a warning on every session start:Impact
Cosmetic only — the four real agents load fine and no skill is affected. But the warning fires on every startup and can be misread by users as "an agent/skill failed to load."
Suggested fix (any one of these)
agents/PERSONAS.mdwon't help (same dir). Instead move it out of the scanned folder, e.g.docs/agents.md, and link to it from the top-levelREADME.md.agents/README.mdso it parses as a (harmless/no-op) entry — not recommended, as it would surface a bogus "agent."README.mdinstead of shipping a README insideagents/.Option 1 (relocate to
docs/) is cleanest and keepsagents/containing only real agent definitions.Environment
agent-skillsplugin v1.0.0