agentsight-bugfix
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill requires the agent to ingest and analyze external bug reports or issue descriptions (Stage 1: "阅读 issue 描述或 Bug 报告"). This content is untrusted and could contain malicious instructions designed to override agent behavior during the debugging cycle.
- Ingestion points: Issue descriptions and bug reports (SKILL.md, Stage 1).
- Boundary markers: None are specified to delimit untrusted data from instructions.
- Capability inventory: File system access (reading code and logs), database querying, and execution of terminal commands including
cargo test,cargo clippy, andagentsight trace. - Sanitization: The workflow does not specify any sanitization, validation, or escaping of the ingested issue data before processing.
- [COMMAND_EXECUTION]: The skill instructions involve executing several shell commands as part of the standard development and verification cycle.
- Evidence: The workflow mandates the use of
cargo fmt,cargo clippy --all-targets -- -D warnings, andcargo testin Stage 4, as well asagentsight tracein Stage 2 and Stage 5.
Audit Metadata