The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes content fetched from arbitrary third-party websites (via Perplexity research and WebFetch) to identify and catalog mentions. This content is untrusted and could contain malicious instructions designed to influence the agent's behavior during the reporting or file update phase.
Ingestion points: External data enters the context through mcp__perplexity__perplexity_research and the WebFetch tool in Step 2.
Boundary markers: Absent; the skill does not specify the use of delimiters or 'ignore' instructions when handling fetched metadata like page titles or snippets.
Capability inventory: The skill possesses file-write capabilities (updating YAML/JSON catalogs and README stats) and shell execution capabilities (running a Python validation script and Git commands).
Sanitization: No explicit sanitization or escaping of external strings is mentioned before they are written to local files or displayed to the user.
[EXTERNAL_DOWNLOADS]: The skill retrieves data from URLs discovered dynamically via search queries. Since these sources are third-party and unknown until runtime, they represent an untrusted ingestion surface for external content.
[COMMAND_EXECUTION]: The workflow involves executing local shell commands, specifically python3 scripts/check-media-mentions.py to validate data integrity and git add/commit to record changes. These commands are triggered by the skill logic and operate on data derived from external sources.

source-command-track-mentions