to-spec
Warn
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: MEDIUMINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONMETADATA_POISONING
Full Analysis
- [METADATA_POISONING]: The skill metadata claims the author is 'Matt Pocock', which contradicts the registered author 'falcondev-oss'. This represents misleading metadata.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from the repository and conversation history, which is then synthesized and used for external write operations.
- Ingestion points: SKILL.md instructions to "Explore the repo" and "take the current conversation context".
- Boundary markers: None present to distinguish untrusted data from instructions during synthesis.
- Capability inventory: SKILL.md instructions to publish content via the
ghcommand-line tool (create, comment, or rewrite issues). - Sanitization: No explicit sanitization or validation of the synthesized content is specified before the automated publication step.
- [COMMAND_EXECUTION]: The skill instructs the agent to use the GitHub CLI (
gh) to perform repository operations such as publishing issues and applying labels.
Audit Metadata