The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates documentation management through several local shell commands and script executions:
Executes a bundled Node.js script, scripts/push-docs.mjs, which aggregates documentation files and pushes them to a user-specified central server using HTTP PUT requests.
Utilizes git diff, git log, and git status to track source code changes and determine which documentation sections require updates.
Uses rg (ripgrep) to perform identifier-based searches across the documentation directory for consistency checks.
[INDIRECT_PROMPT_INJECTION]: The skill generates documentation by processing source code, test files, and changelogs, which presents an attack surface for instructions hidden within codebase comments or string literals.
Ingestion points: Library source code, type definitions, tests, and documentation files such as README.md and CHANGELOG.md as outlined in SKILL.md and change-sync.md.
Boundary markers: The instructions in SKILL.md and doc-standards.md explicitly direct the agent to document only factual exports and strictly adhere to defined formatting schemas to minimize training-data hallucinations and bias.
Capability inventory: The skill possesses the capability to perform network operations (HTTP PUT) via the push-docs.mjs script and write configuration files like .env and .md documentation files.
Sanitization: The documentation transmission script, push-docs.mjs, includes a parseFrontmatter function that validates metadata integrity and ensures mandatory fields like title are present before attempting a push operation.

docs-gen