The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted data from AI search answers and Research Packs, creating a surface for indirect prompt injection.
Ingestion points: Data enters the agent's context through query-corpus.json, Research Packs, and raw answer captures as specified in the Procedure section of SKILL.md.
Boundary markers: The skill employs hash-pinning for artifacts and requires a frozen query corpus to mitigate unauthorized modifications, as outlined in references/measurement-protocol.md.
Capability inventory: The skill possesses the capability to execute shell commands via local Python scripts (validate_ai_visibility_monitor.py, manage_markdown_site.py, manage_markdown_work.py).
Sanitization: Input data is validated against formal JSON schemas (e.g., research-pack.schema.json), providing a layer of structure-based filtering, though specific NL sanitization is not detailed.
[COMMAND_EXECUTION]: The skill utilizes shell-based execution of local Python scripts to validate visibility runs and manage the Markdown-based site inventory.
Evidence: SKILL.md (Step 13) and references/markdown-delivery-contract.md (Shared tools section) instruct the agent to run scripts such as validate_ai_visibility_monitor.py and manage_markdown_site.py via the python command.

ai-visibility-monitor