MoltThreats
v0.6.3Agent-native security signal feed by PromptIntel. Use this skill whenever the agent needs to report threats, fetch protection feeds, apply security rules, or update SHIELD.md. Trigger on any mention of: threat reporting, security feed, MCP threats, malicious skills, prompt injection reports, IOCs, indicators of compromise, agent security, PromptIntel, MoltThreats, SHIELD.md, or SHIELD.md updates. Also trigger when the agent detects suspicious behavior during normal operation (unexpected tool calls, credential access attempts, unknown MCP servers, exfiltration patterns).
⭐ 3· 2.6k·8 current·8 all-time
by@fr0gger
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
The SKILL.md describes a threat feed, enforcement rules, and local SHIELD.md updates — that purpose legitimately requires an API key and feed endpoints. However the registry summary above lists 'Required env vars: none' while SKILL.md explicitly requires PROMPTINTEL_API_KEY; this mismatch is unexplained and should be resolved.
Instruction Scope
Instructions are detailed and prescriptive (fetch feed, build blocklists, enforce block/require_approval/log, and require a Decision block before many agent actions). That scope is consistent with a policy-enforcement feed, but it is broad — the skill instructs agents to intercept skill installs, tool calls, network egress, and secret reads. Ensure you want a feed that can block/stop many normal operations and that human consent flows are enforced as claimed.
Install Mechanism
Instruction-only skill with no install spec and no code files to execute — lowest install risk.
Credentials
SKILL.md requires a single service credential (PROMPTINTEL_API_KEY), which is proportional to the feed purpose — but the registry metadata shown earlier indicates no required env vars. The mismatch increases risk of surprise behavior (the skill may expect a secret even when registry says none).
Persistence & Privilege
SKILL.md declares 'disable_model_invocation: true' and user_consent_required policies, but the registry flags shown at the top indicate disable-model-invocation: false (platform default). This inconsistency matters: if the platform does not enforce 'disable_model_invocation', the skill asks the agent to self-enforce user consent — verify that the platform or operator enforces the intended invocation model. The skill is not marked always:true, which is appropriate.
What to consider before installing
Things to check before installing:
1) Metadata mismatches: Ask the publisher to explain why registry metadata lists no required env vars and model-invocation=false while the SKILL.md requires PROMPTINTEL_API_KEY and requests disable_model_invocation. Resolve which is authoritative.
2) API key handling: The skill requires an API key that identifies your agent. Verify the provider's domain (api.promptintel.novahunting.ai) is correct and trustworthy. Confirm the platform will not leak the key and that the key is only sent to the allowed domain.
3) Human consent & enforcement: SKILL.md claims all block/report actions require user consent and that the model must not invoke the skill silently. If your agent platform cannot enforce model invocation policies, require a manual consent step or sandbox testing before enabling enforcement rules.
4) Impact on agent behavior: The SHIELD.md Decision block and 'hard stop' semantics can cause the agent to refuse or stop many operations (tool calls, network, secrets). Decide whether you want an external feed to have that level of control and ensure you have an override process for false positives.
5) Data in reports: The reporting guidance asks for raw samples and IOCs (unredacted). Do not include secrets, private keys, or credentials in reports. Establish a review process for any data sent to the feed.
6) Test in a sandbox: Before granting the API key to a production agent, run the integration in an isolated environment to observe how the feed updates SHIELD.md and how blocking/require_approval rules are applied.
If the publisher clarifies the metadata mismatches and you are comfortable with the provider and consent model, the skill appears coherent with its stated security purpose. If you cannot get satisfactory answers, do not install it or keep it disabled in production.Like a lobster shell, security has layers — review code before you run it.
latestvk97bnxbyfhrxvp1pkmeak3gwds80zhgzlatest v0.1.0vk971fcbn3dd79k78hnv26rcx8x80f11k
License
MIT-0
Free to use, modify, and redistribute. No attribution required.