flygard-core/.agents
2026-08-27 14:57:14 +02:00
..
docs feat(Core/e2e): live protocol e2e suite (#27158) 2026-08-27 14:57:14 +02:00
skills/generate-pr-description feat(Agents): add generate-pr-description skill (#26586) 2026-07-13 22:38:58 +02:00
README.md docs(Agents): minor improvements (#27211) 2026-08-20 16:00:49 -03:00

.agents

Source of truth for this repo's agent skills and docs:

  • skills/<name>/SKILL.md
  • docs/ — task-scoped agent docs; taxonomy and placement policy in docs/README.md
  • plans/ — per-task planning docs, gitignored

Conventions live in AGENTS.md and docs/. Agent-specific files point there instead of restating them, so there is one copy to keep current.

Hooking up your agent

  • Claude Code — reads CLAUDE.md, which imports AGENTS.md. Skills are exposed by a relative symlink: .claude/skills/<name> -> ../../.agents/skills/<name>.
  • GitHub Copilot — reads AGENTS.md natively. Code review also reads .github/copilot-instructions.md; custom agent profiles live in .github/agents/.
  • Any other agent — point it at AGENTS.md through its own entry file (GEMINI.md, .cursorrules, …), using that agent's include syntax or, where includes are unsupported, a line telling it to read AGENTS.md first.

To add a skill: create it here, then symlink it from each agent's skills dir. Docs need no symlinks — AGENTS.md references them by path.

On Windows, symlinks need git config core.symlinks true plus Developer Mode or an elevated shell; without them git checks the links out as plain text files.

The skills in this repository often build on those from agent-toolkit, invoking them when available. Installing the toolkit at user level is therefore recommended.