Back to skill
Skillv0.1.1
ClawScan security
Skills · ClawHub's context-aware review of the artifact, metadata, and declared behavior.
Scanner verdict
BenignMar 17, 2026, 2:34 AM
- Verdict
- benign
- Confidence
- high
- Model
- gpt-5-mini
- Summary
- The skill's instructions and requirements align with a human-in-the-loop API: it only describes making HTTP calls to sanctifai.com and does not request unrelated system access, though you will need an API key at runtime and should be mindful of what data you send to human workers.
- Guidance
- This skill appears to do what it says: call SanctifAI to create tasks and receive human responses. Before enabling it, consider: (1) you'll need to provide a SanctifAI API key for authenticated operations — prefer a scoped/dedicated key and avoid embedding secrets in URLs when possible, (2) any data you send becomes visible to human workers—do not send sensitive PII/PHI unless you have consent and appropriate controls, (3) if you use webhooks/callback_url you must expose an endpoint and secure it (validate signatures, restrict IPs), and (4) confirm the organizational/privacy implications of creating accounts via API if the skill uses self-registration. If you accept those trade-offs, the skill is internally coherent.
Review Dimensions
- Purpose & Capability
- okName/description (human-in-the-loop tasks) match the SKILL.md: all operations are HTTP calls to SanctifAI endpoints (create_task, list_tasks, wait/webhook). Nothing in the manifest (no binaries, no install) is unexpected for a pure API integration.
- Instruction Scope
- okRuntime instructions only describe calling SanctifAI endpoints, using MCP or REST, long-polling or webhooks, and building forms. There are no instructions to read local files, access unrelated env vars, or exfiltrate data to third parties outside the SanctifAI endpoints. Note: webhooks and callback_url features mean the agent may be asked to accept or provide network endpoints.
- Install Mechanism
- okNo install spec and no code files — lowest-risk model for installation. The skill is instruction-only and does not write code to disk or download third-party packages.
- Credentials
- noteThe manifest declares no required env vars, which is acceptable for an instruction-only skill, but the SKILL.md clearly expects an API key (e.g., sk_live_xxx) for authenticated calls. That credential is proportional to the purpose, but it is not declared up-front — the agent or user will need to supply/store the SanctifAI API key at runtime.
- Persistence & Privilege
- okalways is false and the skill does not request system-wide modifications. Nothing in the SKILL.md asks the agent to modify other skills or global agent settings.