The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources such as ticket titles and test case names, creating a potential surface for instructions embedded in that data to influence agent behavior.
Ingestion points: External data enters the context via cawplan tickets search and qa-insights testrail execution failures in compute-rules.md.
Boundary markers: The instructions do not require the use of explicit delimiters or boundary markers for the ingested content.
Capability inventory: The skill can modify the state of version records through the cawplan qa-insights risk-assessment save command.
Sanitization: There is no guidance on sanitizing or escaping the data retrieved from external sources before interpolation into reports or save payloads.
[DYNAMIC_EXECUTION]: The skill generates data payloads at runtime and stores them in local temporary files for use in subsequent commands.
Evidence: The skill writes a JSON object to /tmp/a5-risk-save.json and then executes the cawplan save tool using the --body-file parameter in save-rules.md.

cawplan-risk-assessment