The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process unstructured application state (user-provided text) through decision models, creating an attack surface where malicious input could attempt to influence the agent's routing or classification logic.
Ingestion points: Untrusted data enters the agent context via the state object described in SKILL.md (Step 4) and the parseRequestBody function in scripts/lib.ts.
Boundary markers: The skill instructions do not require or implement specific delimiters or boundary markers to isolate untrusted input within the model requests.
Capability inventory: The skill utilizes scripts/decide.ts and scripts/models.ts to perform network operations, sending the ingested data to the OpenRouter API.
Sanitization: There is no evidence of explicit sanitization or filtering of external content before it is processed by the decision models.
[COMMAND_EXECUTION]: The skill includes local TypeScript scripts intended to be executed by the user to interact with the API and compare models.
Evidence: README.md and SKILL.md provide commands for running npx tsx scripts/models.ts and npx tsx scripts/decide.ts.
[EXTERNAL_DOWNLOADS]: The skill documentation and scripts utilize external dependencies and network services.
Evidence: scripts/package.json specifies the @openrouter/sdk (a vendor resource) and tsx package. The scripts perform network requests to the openrouter.ai domain, which is the official infrastructure for the skill's service.

openrouter-decisions