Skill Detail

Audit Skills (Premium Universal Security)

This skill appears to be a set of documentation. Driftloom found shell-oriented patterns and destructive commands.

GitHub:sickn33/antigravity-awesome-skills audit-skills
version 669e2578ab7f
static analysis only
no human review yet
High Risk

Current public label

High Risk

The skill includes shell commands that could be dangerous, and the documentation-only structure limits automated verification.

This label is currently coming from the automated scorecard.

Automated result

High Risk

Driftloom found high-severity shell patterns, including a pipe-to-shell install and a hard-delete command. The skill is mostly documentation.

2 high, 3 medium, 1 low Final label: high risk.

Human review

No human review has been recorded yet.

The current public label is still relying on automation. A human has not weighed in yet.

What happened

Driftloom completed a static scan. It inspected the skill files, recorded findings, and generated a scorecard.

Runtime evidence

No sandbox runtime result has been recorded yet.

Driftloom currently recommends runtime testing for this version (priority 35).

What did not happen

  • Driftloom did not run this skill in an isolated sandbox yet.
  • This label is not a guarantee that the skill is safe, bug-free, or appropriate for every environment.
  • A good score does not replace human judgment when a skill touches secrets, shell access, or external systems.

Source provenance

Source: Workspace import

Originally ingested from a local workspace copy.

Scorecard

Safety
0
Quality
94
Transparency
100
Operational
92
Maintenance
44

2 high, 3 medium, 1 low Final label: high risk.

Severity mix: 2 high, 3 medium, 1 low

What Driftloom checked

  • Read the skill files and metadata to understand what the skill claims to do.
  • Looked for shell commands and risky command patterns.
  • Looked for external URLs and network behavior.
  • Looked for secrets and credential handling clues.
  • Checked whether the skill structure and references looked internally consistent.

Findings

Pipe-to-shell pattern detected
shell.curl_pipe_shell · safety
High

Found a curl|sh style install pattern. Very convenient, often very stupid.

File: SKILL.md
Evidence: curl ... | bash
Destructive shell pattern detected
shell.rm_rf · safety
High

Found a hard-delete shell pattern. That deserves human eyes, not optimism.

File: SKILL.md
Evidence: rm -rf
Remote access command referenced
shell.remote_access · safety
Medium

The source references remote access or copy commands.

File: SKILL.md
Evidence: scp
Remote access command referenced
shell.remote_access · safety
Medium

The source references remote access or copy commands.

File: SKILL.md
Evidence: ssh
Privilege escalation command referenced
shell.sudo · safety
Medium

The source references sudo. That may be legitimate, but it changes the risk profile.

File: SKILL.md
Evidence: sudo
Documentation-only skill structure
structure.docs_only · quality
Low

The source looks almost entirely documentation-based, with no obvious code or config files to inspect. That does not make it bad, but it limits how much automation can meaningfully verify.

File: SKILL.md
Driftloom is following the boring, correct architecture: Django control plane first, isolated runner later. Miraculously, restraint survives.