The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data, such as existing user resumes and Job Descriptions (JD), which could potentially harbor malicious instructions.
Ingestion points: SKILL.md (Initial resume parsing and JD-based customization steps).
Boundary markers: references/resume-contract.md establishes a strict "Facts Contract" to ensure agent-provided JD content does not accidentally override or append to candidate facts.
Capability inventory: The skill possesses the ability to write files (resume-facts.yaml, generated HTML/PDFs), execute local scripts for validation, and launch a browser-based editor via npx.
Sanitization: Implements both automated validation through scripts/validate_resume.py and mandatory manual visual review of generated previews.
[COMMAND_EXECUTION]: The skill uses system commands to manage the document rendering and validation pipeline.
Evidence: scripts/render_resume.ps1 executes node to run the rendering logic in render_resume.mjs.
Evidence: tests/test_render_integration.py and tests/test_validate_resume.py utilize subprocess.run to orchestrate integration tests.
Evidence: scripts/validate_resume.py calls node to perform layout measurements using a JavaScript helper.
[EXTERNAL_DOWNLOADS]: Remote code and assets are retrieved during initialization and rendering.
Evidence: SKILL.md directs the agent to execute npx -p @chasen-liao/resume-skills@latest, fetching the vendor's own CLI tools from the NPM registry.
Evidence: HTML templates in references/examples/ fetch font assets from fonts.googleapis.com, a well-known service.
[DYNAMIC_EXECUTION]: The skill employs dynamic loading and runtime code execution.
Evidence: scripts/render_resume.mjs and scripts/measure_resume_layout.mjs use dynamic import() to load the Playwright library.
Evidence: scripts/render_resume.ps1 executes a one-liner via node -e to handle manifest invalidation.
[PRIVILEGE_ESCALATION]: The skill instructs the use of PowerShell with a relaxed execution policy.
Evidence: SKILL.md and references/runtime-setup.md recommend running the renderer with -ExecutionPolicy Bypass to allow local script execution.

resume-builder