The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted content from source code repositories, including potentially attacker-controlled text in GitHub Actions workflows.
Ingestion Points: Target codebase files, configuration files, and GitHub Actions workflow definitions as described in the workflow steps in SKILL.md.
Boundary Markers: No specific delimiters or instructions are provided to the agent to treat target code as data only and ignore any embedded instructions.
Capability Inventory: The agent has the capability to execute shell commands, run Python scripts, and write review reports to the filesystem.
Sanitization: The instructions focus on redacting discovered secrets but do not include measures to sanitize or escape untrusted content to prevent injection during the analysis process.
[COMMAND_EXECUTION]: The skill invokes several external command-line tools to perform security scans.
Evidence: The workflow explicitly instructs the agent to check for and execute tools such as gitleaks, pip-audit, osv-scanner, zizmor, and actionlint in SKILL.md.
[DYNAMIC_EXECUTION]: The skill executes a local Python script for configuration management using a path relative to its own directory.
Evidence: The instruction python3 "${CLAUDE_SKILL_DIR}/../../scripts/fitness-config.py" show --path in SKILL.md executes a script located outside the skill's immediate directory.

review-security