specsfy-specialist-debugging
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and analyze untrusted external data such as error logs, system traces, and metrics to diagnose bugs. This creates a surface for indirect prompt injection where malicious instructions hidden in log files or error messages could attempt to influence the agent's logic during the diagnostic process.
- Ingestion points: The 'Fluxo' section in
SKILL.mddescribes capturing observed behavior, logs, and metrics from the environment. - Boundary markers: There are no explicit instructions to use delimiters or ignore embedded commands within the analyzed data logs.
- Capability inventory: The skill focus is diagnostic, but it references capabilities like Ansible-based collection and interaction with other specialized skills.
- Sanitization: The skill recommends removing sensitive data before concluding, but does not specify input validation or sanitization for the ingestion phase.
Audit Metadata