error-analysis
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill analyzes untrusted trace data from external sources, which could contain malicious instructions designed to influence the agent's analysis or output.
- Ingestion points: API responses or local files loaded into error-analysis/traces.jsonl.
- Capability inventory: Access to Bash, Write, and Agent tools.
- Sanitization: Missing explicit escaping or sanitization of ingested content.
- Boundary markers: Absence of clear delimiters or system instructions to ignore embedded commands in trace data.
- [COMMAND_EXECUTION]: The skill generates shell commands for data retrieval and processing.
- Evidence: Scripts in SKILL.md and references/langfuse-traces.md utilize curl and jq for API interaction.
- Safety context: Includes validation logic to prevent SSRF and credential leaks in URLs, such as enforcing HTTPS for remote hosts and blocking userinfo in URLs.
- [DYNAMIC_EXECUTION]: The skill dynamically constructs a prompt for a sub-agent based on the results of the trace analysis.
- Evidence: The skill uses the Agent tool in Phase 5 to invoke ork:eval-runner with a prompt derived from the generated failure taxonomy and recommendations.
Audit Metadata