The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes internal Python scripts (build_capstone_ppt.py, build_capstone_docx.py) and uses subprocess.run within render_project_diagram.py to call npx. These commands are used to automate the rendering of diagrams and the construction of report documents based on project data.
[EXTERNAL_DOWNLOADS]: The skill relies on npx -y @mermaid-js/mermaid-cli to fetch the Mermaid diagramming tool from the official NPM registry. This is a standard and well-known service for this functionality.
[DATA_EXFILTRATION]: While the skill acknowledges the presence of sensitive files like ~/.peas-agent/config.json, it includes explicit "agent_must_not" instructions and warnings in SKILL.md and references/server-topology.md to ensure that API keys and internal router URLs are never included in the generated reports.
[PROMPT_INJECTION]: The skill is susceptible to indirect injection via student-provided input, though it employs structural safeguards.
Ingestion points: Student project descriptions provided during Steps 1, 2, and 3 of the workflow (processed in references/step-scripts.md).
Boundary markers: The agent is provided with specific templates and instructed to only summarize input into structured lists.
Capability inventory: Subprocess execution for rendering diagrams and file-writing capabilities for generating .md, .pptx, and .docx files.
Sanitization: The workflow includes a mandatory "Alignment Confirmation" step (Step 3c) where the student must verify the summarized data before any files are written to the disk.

peas-capstone-report