The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses dynamic context injection (the !command syntax) in the Context section to execute local shell scripts (get-automation-config.sh and blueprint-autorun.sh) at skill load time to populate the environment.
[COMMAND_EXECUTION]: Employs shell utilities including jq for JSON manipulation of the project manifest and gh (GitHub CLI) for issue management and label creation.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection.
Ingestion points: Reads untrusted data from docs/prps/*.md and the output of gh issue list.
Boundary markers: None identified; instructions do not specify delimiters for external content.
Capability inventory: Uses Bash and the Task tool to dispatch subagents, and gh to write to external repositories.
Sanitization: Absent; the skill explicitly instructs the agent to copy TDD requirements from PRPs "verbatim" into GitHub issue bodies, allowing malicious instructions in PRPs to influence the agent or downstream consumers.

blueprint-autopilot