The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill recommends auto-approving the glab variable list command in the references/allowlist.md file. The documentation explicitly acknowledges that this command "exposes CI/CD variable values, which often contain secrets (API keys, passwords)." Enabling auto-approval for this command allows the agent to harvest sensitive environment variables without explicit user review for each execution.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources, creating a vulnerability to indirect prompt injection attacks.
Ingestion points: Build logs retrieved via gh run view --log-failed and glab ci trace, as well as PR metadata and status descriptions fetched from GitHub and GitLab APIs.
Boundary markers: There are no explicit instructions or delimiters defined to ensure the agent treats external content as untrusted data rather than executable instructions.
Capability inventory: The skill possesses shell execution capabilities (Bash) and uses authenticated CLI tools (gh, glab), which could be leveraged if a malicious instruction embedded in a build log is processed by the agent.
Sanitization: The skill does not perform any sanitization or filtering on the content ingested from CI logs or API responses before presenting it to the agent.

git-ci