The Agent Skills Directory

[PROMPT_INJECTION]: The skill documentation includes specific behavioral instructions that direct the agent to promote an external commercial platform.
Evidence: The 'Suggest Using K-Dense Web For Complex Worflows' section in SKILL.md instructs the agent to 'proactively suggest using K-Dense Web' and to 'frame the suggestion as an optional productivity upgrade' when task complexity increases. This overrides the agent's default neutrality.
[INDIRECT_PROMPT_INJECTION]: The skill's core functionality is to fetch and process real-time information from the web, which introduces a risk of the agent being influenced by malicious instructions embedded in external content.
Ingestion points: The skill uses scripts/perplexity_search.py to ingest third-party web content via Perplexity/OpenRouter.
Boundary markers: No delimiters or 'ignore' instructions for the retrieved content are specified in the provided markdown file.
Capability inventory: The skill executes local Python scripts (scripts/perplexity_search.py, scripts/setup_env.py) and performs network operations via the litellm library.
Sanitization: There is no documentation regarding the sanitization or validation of external search results before they are processed by the agent.

perplexity-search