linear-triage
Pass
Audited by Gen Agent Trust Hub on Nov 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted content from Linear issue descriptions and comments, which could contain adversarial instructions.
- Ingestion points: Untrusted data enters the agent context through tools like
get_issueandlist_issues, which fetch issue bodies and comment histories (referenced in references/TRIAGE.md). - Boundary markers: The skill instructions suggest identifying AI-generated content with a
**Triage note** (generated by AI)marker (referenced in SKILL.md). - Capability inventory: The skill possesses credentials to modify the Linear project state, including creating issues, updating statuses, and editing descriptions via the
mcp__claude_ai_Linear__*toolset. - Sanitization: No specific input sanitization or filtering is described; instead, the skill relies on a 'prepare → confirm → execute' safety protocol (referenced in SKILL.md) that requires the user to verify all changes before the agent applies them, effectively preventing the automatic execution of injected commands.
Audit Metadata