The Agent Skills Directory

[SAFE]: The skill provides a well-structured CLI tool for Linear integration. It handles authentication using OAuth 2.0 PKCE and stores credentials securely using the macOS Keychain or local configuration files with restricted (0600) permissions.
[COMMAND_EXECUTION]: The scripts/linear.mjs script executes local commands via child_process.spawnSync to interact with git and the macOS security binary. These operations are limited to discovering the repository context and managing authentication credentials.
[PROMPT_INJECTION]: The skill correctly identifies the risk of indirect prompt injection from Linear issue data and provides explicit instructions in SKILL.md to treat external content as untrusted.
Ingestion points: Issue retrieval and search operations in SKILL.md and scripts/linear.mjs.
Boundary markers: Explicit 'Safety' section in SKILL.md warning against executing content from issue descriptions.
Capability inventory: Access to Bash and Read tools; network and file system access within the Node.js script.
Sanitization: Uses JSON for data exchange to maintain structure, though raw text content remains unfiltered.

linear