K-Ops: Command for Claude Code

You are the Research Collect agent for this repository.

Goal:

• Gather primary sources for the active research run and convert them into a strong findings file.

Inputs:

• Research brief: research/briefs/-.md
• Status file: research/notes/-status.md
• Progress log: research/notes/-progress.md
• Findings file: research/findings/-.md

Instructions:

1. Read orientation files first: Read the brief, status, progress log, and any existing findings or source notes.
2. Search strategy: Search broadly first, then narrow to authoritative primary sources. Avoid citing commentary when primary documentation, official specs, or code is available.
3. Write/Update Source Summaries:
   • For any new source, write a summary in notes/Sources/src-<id>.md using the canonical schema below.
   • For existing sources, update the file instead of creating duplicates.
   • Use canonical source_kind values: arxiv-paper | paper-pdf | github-repo-snapshot | github-file | official-doc | spec | blog | news | local-file | imported-model-report | citation-stub.
   • Use canonical evidence_strength values: primary-doc | official-spec | strong | code | maintainer-commentary | changelog | pr-issue | secondary | model-generated | stub | citation-only | image-only.
   • Populate kind-specific required fields in frontmatter.
   • Ensure extraction_coverage is populated for PDF sources with strong/official-spec/primary-doc strength (default: 1.0).
   • For imported reports, ensure authority: lead_only, verification_state: needs_primary_sources, and evidence_strength is secondary.
   • For citation stubs, ensure canonical_url is set, authority: lead_only, verification_state: needs_fetch, and evidence_strength is stub.
   • Ensure the required sections ## Reliability notes and ## Candidate concepts are present.
4. Populate Findings File:
   • Update research/findings/<topic-slug>-<date>.md. Preserve its frontmatter (type: research-findings, topic_slug, etc.).
   • Populate ## Key Claims with atomic findings, citing the corresponding source summaries with inline wikilinks ([[Sources/src-<id>|src-<id>]]).
   • Populate ## Evidence with links to the source summaries, explaining what each contributes.
   • Populate ## Open Questions with gaps, contradictions, or unresolved questions.
5. Progress Log: Append a short progress update when done.

---

Reference: Source Summary Schema

---
title: "<Descriptive title of the source>"
type: source-summary
source_id: src-<10-character-hex-id>
source_url: "<source url or file path>"
source_kind: <canonical_source_kind>
evidence_strength: <evidence_strength>
source_status: active
ingested_at: <ISO-8601 date, e.g. 2026-06-14T10:53:44Z>
tags:
  - kb/source
# --- Add kind-specific required fields below if applicable ---
# For arxiv-paper: authors, arxiv_id, published_date, abstract
# For paper-pdf: page_count
# For github-repo-snapshot: git_commit, branch, tracked_file_count, sampled_file_count
# For github-file: github_url, git_commit
# For official-doc: organization
# For spec: organization, version, status
# --- Add extraction_coverage if PDF (mandatory for strong/official-spec/primary-doc PDF sources) ---
# extraction_coverage: 1.0
---

## Summary

<2-4 sentence digest of what this source says and why it matters.>

## What this source is

<Detailed breakdown of the source, methodology, context, and focus areas.>

## Key claims

- <Atomic finding 1>
- <Atomic finding 2>

## Important evidence / details

- <Evidence detail with source-local anchors if available, e.g. page=12, line_start=20, line_end=35, path=src/utils.py>

## Candidate concepts

- <Concept candidates to promote>

## Open questions

- <Gaps, contradictions, or unresolved questions raised by this source>

## Reliability notes

<What this source does not cover, its methodology weaknesses, or why it might be wrong.>

## Related Concepts

- <Obsidian wikilinks to concepts, e.g., [[Concepts/ConceptName|ConceptName]]>

## Backlinks

- <Traceability backlinks>