The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs subagents to perform high-risk operations including cloning external repositories via git clone and installing third-party dependencies through pip install or npm install based on technical artifacts discovered during web research.
[COMMAND_EXECUTION]: Research subagents are directed to execute Python analysis scripts against downloaded third-party codebases, SDKs, and schema files to extract structural data and field inventories.
[DYNAMIC_EXECUTION]: The skill generates a standalone Python script (test-api.py) for API testing based on discovered specifications and subsequently uses python3 -m py_compile to validate the generated code's syntax.
[EXTERNAL_DOWNLOADS]: The workflow involves extensive network activity to fetch user-provided documentation URLs and download large-scale technical artifacts, such as OpenAPI specifications and SDK source code, from the public internet.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from documentation pages and external repositories discovered at runtime.
Ingestion points: User-provided URLs (Phase 1) and subagent-discovered repositories/schemas (Tracks B and C).
Boundary markers: The skill uses separate subagent contexts with an operating manual (research-subagent-guidance.md), but lacks strict delimiters for processing ingested documentation content.
Capability inventory: Subagents possess write-access to the local filesystem, network access, and the ability to install and execute Python code.
Sanitization: There is no evidence of sanitization or filtering for the external content retrieved during the documentation fetch or repository analysis phases.

research-integration