The Agent Skills Directory

[COMMAND_EXECUTION]: The doctor.mjs script utilizes child_process.spawnSync to run several other Node.js scripts located within the .specforge/tools/ directory, such as self-test.mjs and validate-structure.mjs.
[EXTERNAL_DOWNLOADS]: The skill's documentation and SKILL.md suggest using npx @huangrx6/specforge, which downloads and executes code from the NPM registry. This is consistent with the author's identity but involves remote code fetching.
[DATA_EXPOSURE]: The skill includes instructions to scan the entire repository for Markdown documents (Glob 全仓库 Markdown 文档), which are then processed by the AI agent to determine project status and migration paths. This constitutes a surface for indirect prompt injection if malicious content exists in the processed files.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: Tools like instructions.mjs and artifact-graph-status.mjs read and parse YAML and Markdown files from the repository.
Boundary markers: The provided scripts do not appear to use explicit boundary markers or instructions to ignore embedded commands when processing these files.
Capability inventory: The skill has the capability to write files and execute local Node.js commands via spawnSync in doctor.mjs.
Sanitization: Content is processed using regular expressions to extract specific metadata fields, with limited validation of the underlying data structure.

specforge-onboard