The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data sources such as risk registers, defect logs (CSV), and test case management (TCM) exports to generate reports and influence CI gate decisions. This creates a surface where malicious instructions could be embedded in metadata or ticket descriptions to influence the agent's scoring or reporting logic.
Ingestion points: risks.yaml, defects.csv, test-executions.csv, and manual-cases.md (referenced in SKILL.md and references/calibration.md).
Boundary markers: Absent. The instructions do not specify the use of delimiters or warnings to ignore embedded instructions when processing these data sources.
Capability inventory: File system read/write access, terminal command execution (e.g., grep), and execution of local build scripts (e.g., python scripts/build-risk-coverage.py).
Sanitization: Absent. The instructions do not mention sanitizing or validating external content before it is interpolated into generated markdown reports.
[COMMAND_EXECUTION]: The skill provides specific shell command patterns for the agent to follow when scanning repositories for risk tags or running automated coverage builders.
Evidence: Shell command examples for grep -r "risk:R-001" and python scripts/build-risk-coverage.py are provided in SKILL.md (Step 8).

risk-matrix