The Agent Skills Directory

[COMMAND_EXECUTION]: The skill orchestrates the execution of local scripts, specifically scripts/cover_letter.mjs and scripts/build_pdfs.mjs. These scripts handle specialized logic for cover letter approval and PDF conversion, following a structured workflow managed by the agent.\n- [COMMAND_EXECUTION]: Dynamic file path construction is performed for PDF output using variables like 'Company' and 'Job'. The skill explicitly mandates that all illegal characters in these inputs must be normalized to prevent path traversal or other file-system-related injection attacks.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data, including data/resume.md and job.json. It mitigates potential injection attacks through strict 'Ethics Redlines' that forbid fabrication and 'Structure Freezing' rules that prevent the agent from being manipulated into altering the document's core layout. Mandatory Evidence Chain: (1) Ingestion points: data/resume.md, data/profile.md, output/<run_id>/jobs/<job_id>/job.json. (2) Boundary markers: Explicit 'Ethics Redlines' and 'Structure Freezing Rules'. (3) Capability inventory: File writing and local script execution. (4) Sanitization: Filename normalization requirements.

resume-tailor-write