The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted data from primary sources across the internet using the WebFetch and WebSearch tools.
Ingestion points: Primary source documents and web content fetched dynamically during research tasks (SKILL.md).
Boundary markers: While the instructions mandate the use of Evidence Chains and verbatim quotes, they do not include specific delimiters or instructions to ignore commands potentially embedded within fetched documents.
Capability inventory: The agent has broad permissions including Write, Edit, and Glob, enabling it to modify the local file system and internal project structure (bitwize-music-mcp tool) based on its findings.
Sanitization: No explicit sanitization or validation of external content is performed before it is processed by the model or written to storage.
[DYNAMIC_EXECUTION]: The skill implements a pattern for loading instruction overrides from a local file.
Mechanism: The skill calls load_override("research-preferences.md") to fetch and incorporate custom instructions at invocation start.
Impact: This allows the agent's behavior, including verification standards and source priorities, to be dynamically modified by the content of an external file, which could be misused to subvert established research protocols.

researcher