Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
PMP-Agentclaw
v1.0.3AI project management assistant for planning, tracking, and managing projects using industry-standard methodologies. Use when asked to plan projects, track s...
⭐ 0· 615·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
The name/description (project management: EVM, risk scoring, velocity, health checks, WBS, Gantt, RACI) align with the included source, compiled dist files, templates, and CLI commands. The only oddity: SKILL.md frontmatter includes an 'install' entry (id: pmp-agent-install, kind: download) but the registry metadata/install specifications shown to me do not include a download URL or install spec — the package itself contains code and local templates and requires only Node. This is likely a documentation metadata mismatch rather than a capability problem.
Instruction Scope
Runtime instructions (SKILL.md) direct the agent to follow 15 PM rules and to load templates/configs from the skill directory ({baseDir}/configs, {baseDir}/templates). The CLI code reads files the user expressly provides (--file) and resolves a project directory for health checks; there are no instructions to read unrelated system files, environment secrets, or to phone home. Behavior stays within the project-management domain.
Install Mechanism
The package is primarily instruction + included code (dist/ compiled JS present). There is no registry install spec that downloads arbitrary code at runtime. The SKILL.md contains an 'install' metadata item that claims a download install (kind: download) but no download URL or external host is present in the repo or files examined. That inconsistency should be clarified (documentation mismatch) but there is no evidence of a high-risk remote installer or obfuscated download.
Credentials
The skill requests no environment variables or credentials and the code doesn't reference external secrets or cloud credentials. CLI tools accept file paths and project-dir inputs (user-specified), which is expected for a PM utility. No disproportionate secret access is requested.
Persistence & Privilege
Metadata flags: always=false (not force-included), disable-model-invocation=false (normal), user-invocable=true. The skill does not request elevated or persistent system privileges and does not modify other skills' configs. Autonomous invocation is allowed by default on the platform but this skill does not combine that with other high-risk behaviors.
Assessment
This skill appears to implement the advertised project-management features and does not request credentials or perform external network calls. Things to check before installing:
- Verify source/repository: the package.json points to a GitHub URL but the registry metadata shows Source: unknown and Homepage: none. If possible, inspect the repository on GitHub (or obtain the author’s repository) and confirm it matches the package contents.
- Clarify the SKILL.md 'install' metadata: SKILL.md contains an install entry (kind: download) but no download URL or install spec was provided in the registry — ask the publisher whether an external installer is required. Prefer building/installing from the code you reviewed rather than permitting an automatic remote download.
- Run locally in an isolated environment first: build (npm install / npm run build) and run the CLI commands locally (node dist/cli/*) to confirm behavior. This also prevents any surprise network activity from an unreviewed installer.
- Be cautious when supplying file or directory paths to the CLI (--file or projectDir): the CLI will read those files to compute EVM/risks/health — don't point it at system directories or credentials. Only provide project files you intend to analyze.
- If you need higher assurance, review the omitted source files (remaining core/health implementation) and run the code through static scanners locally. Because no scan findings were detected in the provided metadata, that absence is not a guarantee — local verification is recommended.
Overall: coherent and proportionate for a PM assistant, with a small metadata/install inconsistency to resolve and a general recommendation to verify the GitHub source and install from the reviewed code in a sandboxed environment.Like a lobster shell, security has layers — review code before you run it.
latestvk978em38rtz48jfdyh71753nrs81fc6r
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
Runtime requirements
📊 Clawdis
Binsnode
