The Agent Skills Directory

Command Execution: The skill uses the grep utility to scan the docs/integrations/ directory for existing catalog tags. This is an expected administrative task used to ensure that new documentation remains consistent with the repository's taxonomy.
Indirect Prompt Injection Surface: The skill ingests untrusted data from the user (such as product names and descriptions) and interpolates it into Markdown files. This represents a potential surface for indirect prompt injection as external content enters the agent's context.
Ingestion points: User input gathered in Step 1 of SKILL.md (Product name, Category, Tools, etc.).
Boundary markers: None identified in the prompt interpolation process.
Capability inventory: The skill can perform shell commands (grep) and write files to the docs/integrations/ directory.
Sanitization: No explicit sanitization or filtering of user-provided strings is documented. This surface is mitigated by the skill's design, which positions the output as a draft requiring manual review and validation (integration-review and mkdocs serve).
Credential Handling: The skill generates code skeletons for integrations that require authentication. It uses clear placeholders (e.g., YOUR_TOKEN) instead of requesting or hardcoding sensitive credentials, which promotes secure configuration practices among contributors.

integration-create