The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface due to the ingestion of untrusted external data.
Ingestion points: The skill fetches content from external sources including GitHub issues (gh issue view), pull requests (gh pr view), Jira API responses, and local files (memo <path>) as defined in the 'Source Types' section of SKILL.md.
Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions when processing the content of fetched issues or Jira tickets.
Capability inventory: The skill possesses capabilities to write files via mkoutput, execute code through subagents using codex exec, and perform repository modifications via gh pr create.
Sanitization: No evidence of sanitization or validation is present for data ingested from the 'Source Types' before it is interpolated into planning documents or worker tasks.
[COMMAND_EXECUTION]: The skill utilizes command-line tools for automation and subagent management.
Evidence: SKILL.md contains instructions to execute gh (GitHub CLI) for issue and PR management, and codex exec for running code tasks.
Context: The codex exec command is used with the --sandbox flag and subagents are explicitly constrained to READONLY capability, which serves as a security control.

orchestrator