The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs shell command execution and runs local scripts within the repository environment.
It executes python3 .github/actions/gate/run.py --only decisions.adr-fields --dry-run to validate ADR fields using a local tool.
It performs git grep operations using search strings extracted directly from the Decision sections of ADR files.
It executes a shell script to detect and list files in a directory specified by the $POWERSUIT_OVERLAY environment variable.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from documentation files which may influence agent operations.
Ingestion points: Reads the content and frontmatter of all files matching docs/decisions/adr-[0-9]*.md and rfc-[0-9]*.md in SKILL.md.
Boundary markers: The instructions include a "Read-only until the user says go" rule and a strictly defined output format to limit accidental execution, though these do not prevent adversarial injection.
Capability inventory: The skill has the ability to execute python scripts, run git commands, and modify files (ADRs/RFCs) upon user confirmation as seen in SKILL.md.
Sanitization: There is no evidence of sanitization or escaping for strings extracted from ADRs before they are interpolated into git grep commands, presenting a risk of command injection if ADR content is malicious.

decision-log-review