skill-gitea-api.sh now introspects each agent's own PAT (AGENT_TOKEN +
TOKEN_PM/SENIOR/JUNIOR/LEAD/QA) via GET /api/v1/token — a Gitea 1.27+
self-introspection endpoint (token-in, scopes-out, no password) — and bakes a
live, always-accurate 'who can do what' matrix into SKILL.md. Every agent that
loads gitea-api sees every teammate's real scopes; no hand-maintained table to
drift when scopes rotate. Falls back to a static note on <1.27 or unset tokens.
agent.yml passes GITHUB_SERVER_URL + the per-agent TOKEN_* secrets into the
gitea-api skill step so the script can introspect them. run-agent.sh adds a
roster note pointing agents to the matrix for capability-based routing.
Least-privilege gating (plan point #3) was already done in issue #22 via
agents.json skills arrays + install-opencode.sh permission.skill scoping, so
no redundant case gates are added here.
Rebased onto the per-agent skill-scoping change so PR #25 carries both:
- route.sh keeps the registry 'skills' allow-list and emits skills as a step output
- install-opencode.sh writes the permission.skill block (deny-all + allow listed)
Pure refactor otherwise: each step's shell moves to its own file, called via
bash "$SCRIPTS/<name>.sh". The two extracted SKILL.md bodies are byte-identical to
main; routing/config/publish behavior is unchanged. Because this is a reusable
workflow (workflow_call) the runtime checkout is the caller's repo, so agent.yml now
checks THIS repo out into .agents-workflow/ (pinned @main) and points $SCRIPTS there.
senior
lead