Back to skill
Skillv1.0.4

ClawScan security

Macrocosmos · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

SuspiciousFeb 16, 2026, 5:04 PM
Verdict
suspicious
Confidence
medium
Model
gpt-5-mini
Summary
The skill's runtime instructions coherently describe a Macrocosmos API integration that needs an MC_API key, but the registry metadata does not declare that credential (and other metadata mismatches exist), so the package information and runtime requirements are inconsistent.
Guidance
Do not install or supply secrets until the metadata mismatch is resolved. Specific actions to consider: 1) Ask the skill publisher to update registry metadata to declare MC_API as a required secret and to provide the GitHub/PyPI homepage/source links they cite in SKILL.md. 2) Verify the upstream repository (https://github.com/macrocosm-os/macrocosmos-mcp) and PyPI package to ensure the package and endpoint are legitimate. 3) If you must test, create a low-privilege or limited-use MC_API key and avoid using high-privilege credentials. 4) Confirm the domain (constellation.api.cloud.macrocosmos.ai) is the intended recipient of the key and review the provider's privacy/data retention policy. 5) Request an explicit install spec if you expect the Python SDK to be used (so installations are visible/auditable). The current inconsistencies could be innocent (metadata omission) but should be fixed before granting access to secrets.

Review Dimensions

Purpose & Capability
noteThe SKILL.md describes exactly the claimed capability (fetching X/Reddit data via Macrocosmos SN13). The API endpoints, request/response format, and examples align with the stated purpose. However, registry metadata (source/homepage/required env vars) does not match the SKILL.md: the registry lists no required env var or homepage/source, while SKILL.md requires an MC_API key and points to GitHub/PyPI. That metadata mismatch is an incoherence to resolve before trusting the skill.
Instruction Scope
okThe instructions are narrowly scoped to making POST requests to the Macrocosmos SN13 endpoint (and using a Python SDK) and do not instruct the agent to read arbitrary local files or exfiltrate data to unrelated endpoints. Example calls clearly show the API key being used only for requests to constellation.api.cloud.macrocosmos.ai.
Install Mechanism
noteThis is an instruction-only skill with no install spec (lowest install risk). However, SKILL.md references a Python SDK ('macrocosmos' on PyPI) but provides no install guidance. That omission is not necessarily malicious but is an operational inconsistency: an agent may try to install the SDK or fail at runtime if the client is expected but not present.
Credentials
concernSKILL.md requires a secret MC_API environment variable (used as a Bearer token). That credential is proportionate to the skill's function, but the skill registry metadata did not declare any required env vars or primary credential. The missing declaration is a material inconsistency: users may not be warned that the skill needs a secret, and automated permission controls may not be applied. Verify the MC_API scope/permissions before providing it and confirm the registry metadata is corrected.
Persistence & Privilege
okThe skill does not request always:true, does not ask to modify agent/system configuration, and does not request persistent system privileges. Default autonomy (disable-model-invocation:false) is normal and not by itself concerning.