The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes subprocess.call in scripts/setup_student.py to chain the execution of internal utility scripts such as install_skill.py, audit.py, and fix_skills.py during the setup process. Additionally, scripts/audit.py uses subprocess.run to invoke the ledger_report.py script for cost analysis.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted project data including CLAUDE.md, .cursorrules, and other instruction files to calculate token efficiency and detect duplicate content. This creates a surface for indirect prompt injection if these files contain instructions targeting the agent's optimization logic.
Ingestion points: scripts/audit.py (reads rule files and skill metadata) and scripts/compress.py (reads and modifies markdown content).
Boundary markers: The skill does not use explicit delimiters or "ignore instructions" markers when parsing external file content.
Capability inventory: The skill has capabilities to write to files and execute local commands via the subprocess module.
Sanitization: Sanitization is limited to regex patterns used for frontmatter parsing and content classification.
[DYNAMIC_EXECUTION]: The skill performs dynamic execution of its own Python scripts to facilitate its setup and auditing workflows, which are triggered via agent instructions in the provided installation guides.

token-efficiency