debugging-wizard
Pass
Audited by Gen Agent Trust Hub on Sep 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONMETADATA_POISONING
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the analysis of untrusted data sources such as stack traces and error logs, creating a potential surface for indirect prompt injection attacks.\n
- Ingestion points: The workflow involves parsing error messages and correlating log entries, as described in
SKILL.mdandreferences/systematic-debugging.md.\n - Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are implemented to isolate user-supplied error data from the agent's internal instructions.\n
- Capability inventory: The skill utilizes standard debugging and version control tools including
pdb,node,dlv, andgit.\n - Sanitization: The methodology does not include explicit sanitization, validation, or filtering of error logs before they are processed by the agent.\n- [METADATA_POISONING]: The
SKILL.mdfrontmatter contains an author URL that does not match the provided skill author context.\n - Evidence: The skill metadata lists
https://github.com/Jeffallan, while the skill environment identifies the author as1git2clone.\n - Context: The skill documentation includes clear attribution to the original source material (
obra/superpowers), suggesting that the metadata discrepancy is an artifact of forking or templating rather than a malicious attempt to deceive.
Audit Metadata