The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructions involve the agent ingesting external, untrusted data to generate release documentation, creating a potential surface for indirect injection.\n
Ingestion points: The process in SKILL.md reads git commits, RTM.csv, and qa-report.md to produce RELEASE_NOTES.md and CHANGELOG.md.\n
Capability inventory: The skill utilizes the gh CLI (defined in templates/issue-tracker-github.md) to interact with GitHub issues, and it generates/executes local shell scripts for deployment rollbacks (referenced in SKILL.md and templates/prod-checklist.md).\n
Boundary markers: The templates do not currently define explicit delimiters or instructions to ignore potential commands embedded in commit messages or CSV fields.\n
Sanitization: No explicit sanitization or content validation steps are included in the instructions for processing these external inputs.\n- [COMMAND_EXECUTION]: The skill templates and instructions define the use of various development and deployment CLI tools.\n
Evidence: templates/test-plan.md specifies execution of pytest, behave, and playwright. templates/issue-tracker-github.md defines standard interaction patterns using the gh CLI for issue management.\n- [EXTERNAL_DOWNLOADS]: Architecture templates reference an external library for diagram rendering.\n
Evidence: templates/c4.puml uses the !include directive to fetch the C4_Context.puml library from the plantuml-stdlib GitHub repository. This is a well-known community standard for C4 model diagramming in PlantUML.\n- [DYNAMIC_EXECUTION]: The skill incorporates the generation and execution of environment-specific scripts as part of the deployment lifecycle.\n
Evidence: SKILL.md and templates/prod-checklist.md describe creating and testing versioned rollback scripts (e.g., scripts/rollback-<version>.sh) to handle production failures.

ooad-ship