The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local scripts and system commands to gather metadata and manage the research output.
It invokes the script hack/spec_metadata.sh to generate metadata such as researcher name and timestamps.
It uses git branch --show-current, git status, and gh repo view to retrieve repository-specific information for documentation headers.
It performs file system operations, including mkdir -p and writing researchers findings as markdown files into the thoughts/shared/research/ directory.
[INDIRECT_PROMPT_INJECTION]: The skill operates on untrusted data from the codebase, Linear tickets, and user-provided files, which creates a surface for indirect prompt injection.
Ingestion points: The agent is instructed to read codebase files, analyze Linear tickets (via linear-ticket-reader), and process user-mentioned documentation or JSON files (Step 1).
Boundary markers: The skill provides behavioral constraints (e.g., "You and all sub-agents are documentarians, not evaluators"), but it does not use explicit boundary delimiters or "ignore embedded instructions" warnings when processing the contents of external files.
Capability inventory: The agent has the ability to execute local scripts, perform network searches (via web-search-researcher), and write files to the local disk.
Sanitization: There is no evidence of sanitization or escaping of the ingested data before it is synthesized into the final research document.

research