The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes responses from external AI providers (OpenAI, Anthropic, Gemini, Perplexity) which may contain content from untrusted web sources. This text is subsequently analyzed by the agent to compute visibility metrics.
Ingestion points: Raw AI responses, citations, and sub-queries extracted in SKILL.md Step 3 and references/provider-strategies.md.
Boundary markers: The skill uses an 'LLM judge' for disambiguating brand names but lacks specific delimiters or system instructions to ignore instructions embedded in the provider responses.
Capability inventory: The skill writes processed metrics and state updates to MORPHIQ-TRACKER.md and generates a Delta Report JSON.
Sanitization: While the pipeline performs entity normalization and filtering, it does not sanitize retrieved text for instruction-like patterns.
[COMMAND_EXECUTION]: The skill executes local Python scripts to orchestrate the measurement pipeline.
Evidence: Instructions in SKILL.md and scripts/run-queries.py direct the agent to run scripts/create-prompts.py, scripts/diff-results.py, and scripts/generate-report.py for data processing and aggregation.

morphiq-track