The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various local shell commands to introspect the project's build system and codebase status.
Evidence: It runs make -qp and just --list to identify build targets in SKILL.md.
Evidence: It uses grep to scan for TODO comments and debug statements across multiple file types in references/checklist-details.md.
Evidence: It invokes security scanners like gitleaks and trufflehog to search for exposed secrets in the current directory and git history.
Context: These operations are necessary for the skill's primary function of release auditing and are performed locally.
[PROMPT_INJECTION]: The skill contains an attack surface for indirect prompt injection because it reads and processes the contents of user-provided codebase files.
Ingestion points: File contents are ingested via grep searches and security scanning tools (gitleaks, trufflehog) as specified in references/checklist-details.md and references/secrets-scanning.md.
Boundary markers: There are no explicit markers or safety instructions used when presenting the results of these file scans to the agent.
Capability inventory: The skill has the ability to execute significant local tools (make, npm, git, cargo), creating a risk if an attacker embeds instructions within the repository files.
Sanitization: No sanitization or filtering of the file contents or scanner outputs is performed before analysis.

release-readiness-checklist