ClawSec Advisory Feed
Keep OpenClaw agents informed about emerging security threats.
The documentation discloses installation downloads, optional feed updates, no outbound data submission, and no standalone persistence, and it describes signature/hash checks plus traversal and file-count limits. However, the actual deployment flow does not verify the signed manifest, dependencies are incompletely declared, installation writes to the user directory, and user confirmation, isolation, and rollback are absent, so points are deducted.
Installation and feed examples include retries, failure checks, JSON validation, and error messages, making the happy path plausible. Reliability is reduced by undeclared commands, inconsistency between the signed-verification procedure and the deployment procedure, platform-specific date/tool behavior, and the incomplete state-management material. Static assessment caps this dimension at10, so it remains at7.
The target scenario, OpenClaw integration, trigger terms, feed schema, and non-automation boundary are mostly clear. However, the skill itself is primarily an advisory feed while README claims such as real-time monitoring and automatic cross-reference depend on clawsec-suite; non-OpenClaw boundaries and Chinese support are not explained, and core downloads/updates depend on GitHub and other overseas endpoints, warranting deductions.
The candidate includes version 0.0.11, skill.json metadata, a changelog, AGPL-3.0-or-later licensing, installation guidance, and maintenance/reporting entry points. The publisher is not FollowSkills-verified, maintenance ownership and update commitments are not specific, and README and SKILL.md differ in how they describe automation, so the score is limited.
It offers feed retrieval, JSON parsing, severity/exploitability filtering, installed-skill matching, and notification formats, so the core task is plausible. However, standalone mode does not provide continuous polling, automation requires the separate suite, and feed contents, matching accuracy, and directly usable outcomes are not verified in the supplied files. Static calibration caps this dimension at7, so it receives5.
There are versioned documents, example JSON, a signed-manifest verification procedure, and repository CI/test signals. The cited tests mainly cover the suite or other consumers rather than this skill's key paths, and nothing was executed or independently corroborated in this review. Static calibration caps this dimension at5, so it receives4.
- Do not interpret README claims about real-time monitoring, automatic cross-reference, or continuous polling as capabilities of the standalone skill; they depend on clawsec-suite or an external scheduler.
- Before production installation, ensure the signed manifest is actually verified rather than relying only on checksums.json structure and file hashes; also declare runtime tools such as openssl, grep, find, stat, date, cp, and mktemp.
- GitHub and other overseas endpoints are core installation and update paths; mainland-China network restrictions may break monitoring and updates, so an auditable mirror or offline process is needed.
- The skill downloads and writes under ~/.openclaw without explicit confirmation, backup, or rollback steps; review and test it in an isolated environment first.
What does this skill do, and when should you use it?
ClawSec Feed is an advisory-monitoring skill for OpenClaw-related threats and vulnerabilities. It consumes a consolidated feed containing NVD CVEs, approved community advisories, and provisional GitHub Security Advisories. The skill can filter and prioritize advisories, compare affected entries with locally installed skills, and track previously seen advisories in a state file. It is suitable as a lightweight standalone feed or as a component of ClawSec Suite.
Uses curl to download release metadata, signed manifests, artifacts, individual files, and the advisory feed; uses jq to validate and query JSON; verifies release signatures, public-key fingerprints, SHA-256 checksums, archive contents, file counts, and feed structure; scans the local skills directory and matches installed skill names against advisory affected lists; filters and sorts advisories by severity, recency, and exploitability_score; and records check timestamps, feed timestamps, and known advisory IDs in ~/.openclaw/clawsec-feed-state.json. A standalone install writes local skill files only, does not create cron jobs, and does not submit data externally.
- An OpenClaw operator wants recurring checks for malicious skills, vulnerable versions, and prompt-injection patterns.
- A user needs threat intelligence without installing the rest of ClawSec Suite.
- A security maintainer wants to cross-reference published advisories with locally installed skills.
- An operator needs to verify release signatures and checksums before installing a skill on production hosts.
- A team wants to prioritize threats using exploitability context in addition to severity.
What are this skill's strengths and limitations?
- Combines NVD CVEs, community advisories, and provisional GitHub advisories in one feed.
- Includes signed-manifest, public-key, archive, and file-integrity verification procedures.
- Supports severity and exploitability prioritization plus local skill cross-referencing.
- Can be installed independently or deployed from ClawSec Suite.
- The skill targets OpenClaw; other platforms require adaptation.
- Standalone installation depends on several Unix commands and network access.
- The skill does not create scheduled jobs, so recurring monitoring needs the suite or an external scheduler.
- The supplied material does not document a standalone test suite or offline feed mode.
How do you install this skill?
Install with the Vercel Skills CLI: npx skills add prompt-security/clawsec --skill clawsec-feed -a openclaw -y. For standalone deployment, save SKILL.md under ~/.openclaw/skills/clawsec-feed/ and follow the documented release-download workflow. Before production installation, verify the signed checksums.json manifest, public-key fingerprint, archive checksum, and individual file checksums. Alternatively, install clawsec-suite, which can deploy its bundled copy.
How do you use this skill?
Add the feed to a heartbeat routine: check for new advisories, cross-reference installed skills, and notify the user when action is required. Fetch the feed with curl -sSL --fail --show-error --retry 3 --retry-delay 1 "$FEED_URL", then use jq to inspect counts, severity, publication dates, or exploitability_score. Set CLAWSEC_FEED_URL for a custom feed and CLAWSEC_INSTALL_DIR for a custom installation directory. The documented heartbeat interval is 15-30 minutes, with a five-minute minimum; automation must be provided by clawsec-suite or an operator-managed scheduler.
How does this skill compare with similar options?
Compared with ClawSec Suite, clawsec-feed focuses on advisory retrieval, parsing, cross-referencing, and release verification rather than the suite's other security skills. It can run independently or be deployed as a bundled suite component.