Agent Skills: Clous Knowledge Retrieval

Use when retrieving Clous HR knowledge, support docs, or system log context through Clous tool endpoints or MCP resources.

UncategorizedID: clous-ai/agents/knowledge-retrieval

Repository

clous-aiLicense: NOASSERTION

Install this agent skill to your local

pnpm dlx add-skill https://github.com/clous-ai/plugins/tree/HEAD/clous-use/skills/knowledge-retrieval

Skill Files

Browse the full folder contents for knowledge-retrieval.

Download Skill

Loading file tree…

clous-use/skills/knowledge-retrieval/SKILL.md

Skill Metadata

Name
knowledge-retrieval
Description
Use when retrieving Clous HR knowledge, support docs, or system log context through Clous tool endpoints or MCP resources.

Clous Knowledge Retrieval

Use knowledge tools when the user asks for policy, support, HR guidance, or operational diagnostics grounded in Clous data.

Sources

  • /v1/tools/hr-knowledge-retrieve/
  • /v1/tools/clous-support-docs-retrieve/
  • /v1/tools/systemlog-errors-retrieve/
  • MCP resources when connected.

Output Rule

Report source, confidence, and unresolved gaps. Do not present retrieved snippets as legal advice.

Retrieval Contract

  1. Start with the user's operating question and choose the narrowest retrieval source.
  2. Quote or summarize only the minimum context needed for the decision.
  3. Include the source surface, document/resource id when available, retrieval time window, and confidence.
  4. Separate retrieved facts from your interpretation.
  5. If the answer affects HR, performance, legal, compensation, or employment status, mark unresolved gaps and escalate to the relevant owner.

Knowledge Boundary

  • Treat maintained knowledge pages, current HR policy, and product-owned docs as canonical over raw intake, logs, support snippets, and session notes.
  • Use raw sources only as evidence with provenance; do not promote gated, deleted, metadata-only, or unreadable sources into claims.
  • When canonical knowledge and raw evidence disagree, report the conflict, cite both surfaces, and ask the content owner to reconcile.

Guardrails

  • Do not merge stale support docs with current HR policy without labeling the date gap.
  • Do not infer legal, disciplinary, or medical conclusions from snippets.
  • Do not expose individual employee details when an aggregate answer is enough.