audit-noise
Pass
Audited by Gen Agent Trust Hub on Jul 30, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes untrusted data from markdown files within the repository, creating a surface for indirect prompt injection where malicious instructions in a target file could influence the agent's behavior during the audit process.
- Ingestion points: The
scripts/detect.shscript reads the content of arbitrary.mdfiles and the.claude/topic-docs.yamlconfiguration file within the repository. - Boundary markers: The skill respects
<!-- markdown-discipline-ignore -->tags to skip content, but does not use explicit boundary markers or delimiters when presenting excerpts to the agent in the final output table. - Capability inventory: The skill is strictly read-only and is restricted to shell tools (
bash,git,grep,awk,find). It lacks capabilities for network access, file writing, or credential access. - Sanitization: Findings include excerpts that are limited to 120 characters by the
audit_noise_trim_excerptfunction inscripts/lib/noise-shapes.sh.
Audit Metadata