The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (SAFE): The skill performs legitimate API calls to plausible.io to fetch analytics data as documented in lib/client/plausible.ts.
[INDIRECT_PROMPT_INJECTION] (LOW): The skill's architecture for processing external data creates a potential injection surface. 1. Ingestion points: External analytics data from the Plausible API (lib/client/plausible.ts) and web page content fetched via sub-agent prompts as instructed in recipes/comprehensive-audit.json. 2. Boundary markers: Absent. There are no clear delimiters or instructions for the agent to disregard instructions within the data it analyzes. 3. Capability inventory: Network operations (fetch), local file writing for caching (lib/utils/cache.ts), and CLI execution via bun or tsx. 4. Sanitization: Structural Zod validation is applied to API responses, but no semantic sanitization is performed on external content analyzed by the agent.
[DATA_EXPOSURE] (SAFE): The skill accesses the file system to maintain a cache at ~/.cache/plausible-cli and a log at ~/.plausible-agent.log. These are restricted to the user's home directory and serve standard operational roles.
[PROMPT_INJECTION] (SAFE): Instructions in SKILL.md and prompts in the recipes/ directory are benign and strictly follow the skill's stated SEO consultancy purpose.

plausible-insights