K Deep Research
ReviewAudited by ClawScan on May 10, 2026.
Overview
The skill is mostly an instruction-only research workflow, but it should be reviewed because it directs agents to use broad local Obsidian/knowledge files without clear scoping and its embedded metadata conflicts with the registry summary.
Install only if you want an exhaustive research workflow. Before use, set clear rules for which local folders or Obsidian vaults the agent may read, keep secrets/private notes out of scope, and review the full referenced files and validation script if available. Use workspace-scoped installation first rather than making it shared across all agents.
Findings (3)
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.
Private notes or prior research could be pulled into the agent's context or generated reports without the user realizing how much local material was consulted.
The skill instructs agents to use broad local knowledge sources as part of research, but the artifacts do not define allowed paths, exclusions, approval gates, retention, or limits on reuse of private notes.
Filesystem (existing knowledge): - Obsidian vault (4000+ files) - Prior investigation notes, timelines, frameworks
Before enabling, define exactly which vault/folders may be read, exclude secrets and private journals, and require confirmation before local-note content is included in outputs or follow-up research.
Users may not get an accurate registry-level view of when the skill is eligible or what local/network tooling it expects.
The embedded skill metadata declares a curl/wget dependency and alwaysActive=true, while the supplied registry metadata reports no required binaries and always=false, making activation and dependency expectations inconsistent.
metadata: { "openclaw": { "emoji": "🔬", "requires": { "binsOneOf": ["curl", "wget"] }, "alwaysActive": true } }Verify the installed package metadata directly, and prefer workspace-scoped installation until the dependency and activation settings are clarified.
If a user copies these examples into cron/heartbeat configuration, the agent could perform recurring research and send notifications without a per-run prompt.
The reference material includes examples for recurring autonomous monitoring, notifications, and memory maintenance. These appear to be documentation examples rather than automatically enabled behavior.
"Research Monitoring Cron" ... "Check monitored topics in TASKS.md for new developments. If significant findings, write report to research/ and notify via telegram."
Treat the autonomy examples as optional templates only; add explicit schedules, budgets, notification rules, and human approval before enabling any recurring job.