The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external data from Yelp, including user-generated reviews and public business pages, which could contain malicious instructions.
Ingestion points: Data enters the agent's context through Yelp API endpoints (such as api.yelp.com/v3/businesses/{id}/reviews) and public business listings on www.yelp.com as described in SKILL.md and review-analysis.md.
Boundary markers: Present. The review-analysis.md and access-boundaries.md files provide clear instructions for the agent to separate evidence from inference, disclose data sources (API vs. page), and mark uncertainty in findings.
Capability inventory: The skill uses curl and jq for data retrieval and processing, and has the ability to write state and logs to ~/Clawic/data/yelp/ as specified in api-workflows.md and memory-template.md.
Sanitization: The skill includes operational guidelines in review-analysis.md to identify low-confidence situations (e.g., low review volume, stale data) and requires specific signal weighting to mitigate the impact of anomalous or malicious review content.
[COMMAND_EXECUTION]: The skill provides functional templates for shell commands to interact with the Yelp API.
Evidence: api-workflows.md contains multiple curl command patterns used for business discovery, phone number resolution, and review retrieval.
Context: These commands are restricted to official Yelp API domains and include explicit logging safety rules in SKILL.md and api-workflows.md to prevent the leakage of bearer tokens or raw request headers into local memory files.

Yelp