implement-issue-tree
Pass
Audited by Gen Agent Trust Hub on Aug 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from GitHub issue titles, descriptions, and pull request comments, which creates a potential surface for indirect prompt injection attacks.
- Ingestion points: External content enters the agent context during the tree construction phase (
SKILL.md, Step 1), implementation planning (SKILL.md, Step 3), and during continuous monitoring of PR reviews (SKILL.md, Step 6). - Boundary markers: The skill employs isolation techniques, including
<untrusted-data>tags and dynamic nonces (UNTRUSTED_<nonce>_BEGIN) generated from a cryptographically secure seed to encapsulate external data. - Capability inventory: The skill utilizes
ghandgitCLI tools to perform write operations, such as committing code, pushing branches, and merging pull requests. - Sanitization: Comprehensive input validation, escaping of control characters, and architectural context separation (ensuring agents that read untrusted text do not possess destructive permissions) are implemented to mitigate the risk of malicious input influencing agent behavior.
Audit Metadata