The Agent Skills Directory

[PROMPT_INJECTION]: Indirect prompt injection surface. The skill's core optimization loop involves reading external SKILL.md files and executing them within live tests (Phases 1 and 2), creating a potential path for malicious instructions to influence the agent.
Ingestion points: The skill reads all instructional files matching the pattern .claude/skills/*/SKILL.md into the agent's context.
Boundary markers: There are no explicit boundary markers or instructions to ignore embedded commands when these external skills are interpolated into the test environment.
Capability inventory: The skill possesses extensive capabilities, including executing shell commands via git and Node.js, writing to the local filesystem, and spawning autonomous sub-agents.
Sanitization: The skill does not perform sanitization, validation, or escaping of the ingested content before it is executed during the effectiveness evaluation phase.
[COMMAND_EXECUTION]: The skill frequently executes shell commands to perform version control operations and generate visualization assets.
Evidence: Instructions in SKILL.md direct the agent to use git for branching, committing, and reverting changes, and to run node scripts/screenshot.mjs to generate report images.
Script Analysis: The screenshot.mjs script uses execSync to call the macOS open command and contains environment-specific hardcoded paths (e.g., /Users/alchain/...).

darwin-skill