atlas

Fail

Audited by Snyk on Nov 22, 2026

Risk Level: CRITICAL
Full Analysis

CRITICAL E004: Prompt injection detected in skill instructions.

  • Potential prompt injection detected (medium risk: 0.30). The skill is a legitimate state-machine orchestrator, but it silently appends .atlas/ to the user's .git/info/exclude without asking or reporting that change in the OUTPUT summary, which alters the user's repository metadata beyond what the task needs.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

  • Third-party content exposure detected (low risk: 0.10). The required workflow ingests third-party skill bodies from the extracted skills/ tree (ranked by skillselect and injected as the ACTIVE skill into the coder packet), which is a community-sourced package set the agent does not explicitly select, but the ranking is intent-driven and not forced delivery, so this is low rather than medium.

MEDIUM W013: Attempt to modify system services in skill instructions.

  • Attempt to modify system services in skill instructions detected (low risk: 0.10). Operational change gate is met (it appends to the repo's .git/info/exclude, creates git worktrees, and runs a git reset --hard rollback), but the irreversible-loss gate fails because the reset targets a sanctioned isolated worktree with committed green checkpoints and no unrecoverable loss of operational state is shown.

Issues (3)

E004
CRITICAL

Prompt injection detected in skill instructions.

W011
MEDIUM

Third-party content exposure detected (indirect prompt injection risk).

W013
MEDIUM

Attempt to modify system services in skill instructions.

Audit Metadata
Risk Level
CRITICAL
Analyzed
Nov 22, 2026, 01:14 AM
Issues
3