The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data such as library documentation and source code to generate or update agent instructions.
Ingestion points: Ingests library READMEs, Git diffs, and documentation files during the domain discovery and skill generation phases.
Boundary markers: Utilizes mandatory 'STOP gates' and interactive maintainer interviews between analysis phases to mitigate the risk of autonomous obedience to embedded instructions.
Capability inventory: The skill guides the use of CLI tools that perform file system writes (SKILL.md, package.json, GitHub workflows) and network operations (via gh CLI and package managers).
Sanitization: Relies on structural validation (e.g., intent validate) and maintainer review of generated content rather than automated sanitization of ingested text.
[COMMAND_EXECUTION]: The guides instruct the agent to execute various CLI commands for repository maintenance, including modifying configuration files, installing agent hooks, and performing Git-based reviews.
[EXTERNAL_DOWNLOADS]: Fetches configuration templates, package archives, and executable guidance from official TanStack GitHub repositories and the npm registry. These operations target well-known and trusted sources associated with the package vendor.

tanstack-intent