accesslint-audit

Pass

Audited by Gen Agent Trust Hub on Jul 9, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it establishes a trust chain where external tool output directly dictates file system modifications.
  • Ingestion points: The skill ingests untrusted data from URLs, local files, and HTML strings using the audit_live and audit_html flows (SKILL.md).
  • Boundary markers: There are no instructions provided to wrap tool outputs in delimiters or to ignore potential instructions embedded within the audited content (SKILL.md).
  • Capability inventory: The skill possesses the capability to read files, perform network requests to live URLs, and perform write/edit operations on the codebase during 'Fix mode' (SKILL.md).
  • Sanitization: The instructions explicitly discourage human review of mechanical fixes, stating that 'Fix:' fields are 'authoritative' and should be applied 'verbatim' (SKILL.md).
  • [COMMAND_EXECUTION]: The skill implements an automated 'Fix mode' which involves a high-privilege loop of auditing, editing, and verifying files within the user's codebase.
  • Evidence: The skill description and recipe define an 'audit→edit→verify loop on a target' and instruct the agent to 'Group same-file edits into one operation' (SKILL.md).
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 9, 2026, 07:07 AM
Security Audit — agent-trust-hub — accesslint-audit