The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local scripts such as linear-api.sh and sync-github-issues.sh, along with standard CLI tools like git and gh. It references a PreToolUse hook designed to auto-approve specific Bash command patterns, which reduces manual verification of executed operations.
[COMMAND_EXECUTION]: Shell commands are dynamically assembled at runtime. GraphQL queries and mutations are constructed using printf and string interpolation of variables that may contain data fetched from external sources.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted data from Linear issue titles, descriptions, and git commit logs to draft PR bodies and issue comments.
Ingestion points: Data enters the context via [LINEAR-DIGEST] hook context, Linear API responses, and git log output.
Boundary markers: No specific delimiters or instructions are used to isolate untrusted data when it is interpolated into prompts or PR templates.
Capability inventory: The skill can execute subprocesses via bash linear-api.sh and perform file writes to ~/.claude/linear-sync/state.json and .claude/linear-sync.json.
Sanitization: While the skill uses printf for query structure, there is no evidence of sanitization or escaping for the natural language content extracted from issues or logs before it is processed by the agent.

linear-sync