The Agent Skills Directory

[PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill methodology requires the agent to read and mirror content from project documents and repository configuration files (e.g., CLAUDE.md, AGENTS.md), creating a surface for indirect prompt injection.
Ingestion points: The agent is directed to process configuration files and documentation within the docs/ directory to derive Linear structure.
Boundary markers: The skill does not implement explicit delimiters for ingested text, though the methodology instructs the agent to treat document claims as evidence that requires corroboration.
Capability inventory: The agent possesses full management capabilities over Linear projects via MCP tools and the ability to execute shell-based HTTP requests for asset uploads.
Sanitization: The instructions rely on agent-level verification of evidence (e.g., git commits) rather than automated content sanitization.
[COMMAND_EXECUTION]: Integration Command Usage. The methodology validates the use of shell commands such as curl to perform PUT requests to signed URLs for file attachments. This is a standard API interaction pattern for Linear, though it involves the agent executing shell-level operations.

linear-methodology