First-Principles-Deconstructor
PassAudited by ClawScan on May 1, 2026.
Overview
This is an instruction-only analytical skill with no code or credentials, though it does ask the agent to use a scoped local handoff file and write a local report.
This skill appears safe to install as an instruction-only research/audit helper. Be aware that it may read ~/.openclaw/swarm_tmp/expert_output.json and write ~/.openclaw/swarm_tmp/audit_report.json, so clear or review those files if you do not want prior agent outputs to affect the audit.
Findings (2)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
If another process or prior task writes misleading content to that file, the skill may reflect it in the final audit.
The skill incorporates a persisted local JSON file into its analysis. This is scoped and purpose-aligned, but the file’s origin and trust level are not defined, so its contents could influence the audit.
Consume data from `~/.openclaw/swarm_tmp/expert_output.json` if available.
Use this skill when you are comfortable with it reading that local handoff file, and clear or inspect the file if prior task context should not influence the result.
Shared agent output may be treated as part of the audit without clear provenance checks.
The wording indicates a file-based handoff from another expert or swarm step, but the artifact does not define how that producer is identified or validated.
Phase 4: Chaining & Cross-Verification (Critical)
Only use the chaining feature with trusted prior outputs, or review the expert_output.json file before relying on the generated report.