The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute several shell commands for testing, building, and auditing manuals, including node site/build-manuals.js, node manuals/_shared/figkit.js, and node scripts/audit_manuals.js. These operations are presented as part of a standard development lifecycle within the 'AI Engineering from Scratch' repository.
[DYNAMIC_EXECUTION]: The workflow relies on a 'capture kit'—a script provided within the documentation project's directory (capture/run.py). The skill requires the agent to run this script using python3 to generate or verify technical captures, which constitutes execution of code provided alongside the data being documented.
[EXTERNAL_DOWNLOADS]: The audit and build process includes the command npx --yes pagedjs-cli@0.4.3, which downloads and runs the pagedjs-cli package from the NPM registry to generate PDF output.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external sources such as specifications, source code, and log files, presenting an indirect prompt injection surface.
Ingestion points: External content is read from the research/sources/ and capture/out/ directories during the manual creation and fact-checking process.
Boundary markers: The system utilizes structured Markdown fence blocks (listing, rule) with specific metadata fields to distinguish external technical data from the manual's primary prose.
Capability inventory: The execution environment permits shell access and script execution (Node.js, Python, and NPX) for the build and validation steps.
Sanitization: An 'audit' script is described that enforces linguistic rules and structural integrity (Simplified Technical English), but it does not specify security-focused sanitization or filtering for potentially malicious instructions embedded in the ingested data.

write-manual