The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides the agent with the authority to execute deployment steps and create verification tests, which is necessary for its stated purpose.
Evidence: SKILL.md instructs the agent to "Execute each step — or provide commands to execute" and "Create ad-hoc verification tests if needed".
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources such as CI/CD logs and pull request statuses, which presents a surface for indirect instructions to influence agent behavior.
Ingestion points: The agent reads "CI/CD Context", "pipeline" status, and "user behavior" (as noted in SKILL.md and references/reality-contact.md).
Boundary markers: No explicit delimiters or instructions to ignore embedded commands within the external data were found.
Capability inventory: The agent can label PRs, push to deploy branches, execute shell commands, and generate/run ad-hoc test scripts (SKILL.md).
Sanitization: No evidence of sanitization or validation of the ingested external content was observed.

ship