The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it automates release gate checklists by ingesting and summarizing untrusted external data.
Ingestion points: According to modules/quality-signals.md and modules/deployment-readiness.md, the skill reads from GitHub API responses (check suite status, deployment outcomes, and issue labels), Pull Request descriptions, and external tracker CSV exports.
Boundary markers: The instructions lack explicit delimiters or boundary markers to ensure the agent ignores potentially malicious instructions embedded within the ingested data.
Capability inventory: The skill is designed to write comments to deployment PRs, update issues, and modify local tracker documentation, which could be leveraged if the agent follows instructions injected into a PR description or issue label.
Sanitization: No sanitization or validation routines are specified for the data ingested from GitHub or tracker exports before it is used to generate reports.
[SAFE]: The skill performs legitimate operations with GitHub's official API to collect quality signals.
It utilizes standard endpoints such as /repos/{owner}/{repo}/commits/{sha}/check-suites and /repos/{owner}/{repo}/deployments to verify release readiness, which is consistent with its stated purpose of release health governance.

release-health-gates