The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to interact with the Linear GraphQL API and repository data, creating a surface where malicious content from external issues or comments could potentially influence agent behavior during implementation or validation tasks.
Ingestion points: External data enters the context via the WebFetch tool for API responses and Read/Grep for local repository evidence (SKILL.md, references/official-docs.md).
Boundary markers: The skill instructs the agent to use "recorded, redacted GraphQL fixtures" and "redacted receipts," though it does not define specific prompt delimiters for live data ingestion.
Capability inventory: The skill is granted WebFetch, Write, and Edit permissions, allowing it to interact with both remote APIs and the local filesystem.
Sanitization: Explicit instructions are provided to redact credentials, customer content, and sensitive GraphQL variables before writing receipts or files.
[SAFE]: The skill employs WebFetch to access official Linear documentation and API endpoints (linear.app). As Linear is a well-known service, these network operations for documentation and standard API interaction are documented neutrally and do not escalate the risk profile.
[SAFE]: The skill establishes rigorous guardrails against credential exposure, instructing the agent to never write secrets to files and to gate production mutations behind explicit owner approval.

linear-ci-integration