Review and consolidate project findings to ensure a coherent, trustworthy knowledge base.
Install
mkdir -p .claude/skills/consolidate-knowledge-base && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/15042" && unzip -o skill.zip -d .claude/skills/consolidate-knowledge-base && rm skill.zipInstalls to .claude/skills/consolidate-knowledge-base
Activation
This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.
Interactively review the OKF knowledge base to detect contradictions between Findings, mark contradicted/superseded Findings, propose Experiments to resolve open questions, and propose promotions of converged Findings into stable Concept/Structure/Principle documents. Every mutation is confirmed by a human before it is written. Use periodically or after a batch of recorded Findings.Key capabilities
- →Detect contradictions between Findings in the knowledge base
- →Mark Findings as contradicted or superseded
- →Propose Experiments to resolve open questions
- →Propose promotions of converged Findings to stable documents
- →Scan knowledge base files for claims, tags, and links
- →Append lifecycle frontmatter to Findings
How it works
The skill interactively reviews the knowledge base, identifies contradictions, and proposes changes like marking findings or promoting insights, with every mutation requiring human confirmation before writing.
Inputs & outputs
When to use consolidate-knowledge-base
- →Reviewing contradictions after a batch of new findings
- →Updating findings with superseded or contradicted status
- →Promoting validated findings to stable principles
About this skill
consolidate-knowledge-base
Turn a pile of dated Findings into consolidated, trustworthy knowledge — without ever rewriting history. This skill is interactive: it proposes, the human confirms, then it writes.
When to use
- After a batch of Findings has been dumped by
record-finding. - Periodically, to keep the empirical layer coherent with the stable layer.
- When you suspect two findings disagree.
Guardrails
- NEVER edit or delete a Finding's body. Only append lifecycle frontmatter
(
status,contradicted_by,superseded_by). - NEVER promote to a Principle without explicit human confirmation — principles are agreed with humans.
- Ask before every write, using
vscode_askQuestions.
Procedure
1. Scan
Read knowledge/findings/, plus principles/, structures/, concepts/.
Build a map of claims, tags, and existing contradicts/supersedes links.
2. Detect contradictions
Identify pairs where a newer Finding disagrees with an older Finding, or where a Finding conflicts with a Principle/Structure/Concept. Present each candidate contradiction to the human for confirmation.
3. Mark (on confirmation)
For each confirmed contradiction, append metadata-only fields:
- On the newer finding:
contradicts: [<old-id>](if not already present). - On the older finding:
status: contradictedandcontradicted_by: [<new-id>]. Use the same pattern withsupersedes/superseded_byfor replacements. The body of both files stays untouched.
4. Propose Experiments
For unresolved or low-confidence contradictions, propose a new Experiment
document (hypothesis, steps, expected_signals, max_runs: 1–2) that would
settle the question. Write it to knowledge/experiments/ only on confirmation.
5. Propose promotions
When several Findings converge on a stable truth, propose promoting the insight into the stable layer:
- "how it works" →
Structureinstructures/ - "the idea" →
Conceptinconcepts/ - "the standard/convention" →
Principleinprinciples/(always human-agreed) Link the promoted doc back to its supporting findings (supported_by/related_concepts). Write only on confirmation. Findings remain untouched.
6. Validate
just knowledge-lint
just knowledge-validate
Both must pass before finishing.
Output
A short chat summary: contradictions found and marked, experiments proposed, promotions made, and any open questions left for a future run.
When not to use it
- →When the goal is to edit or delete a Finding's body
- →When promoting to a Principle without explicit human confirmation
Limitations
- →The skill never edits or deletes a Finding's body
- →Promotion to a Principle requires explicit human confirmation
- →The skill requires human confirmation for every write operation
How it compares
This skill provides a structured, human-confirmed process for knowledge base consolidation, ensuring data integrity and preventing accidental overwrites, unlike manual updates.
Compared to similar skills
consolidate-knowledge-base side by side with the closest alternatives in the catalog.
| Skill | Installs | Updated | Safety | Difficulty |
|---|---|---|---|---|
| consolidate-knowledge-base (this skill) | 0 | 1mo | Review | Intermediate |
| notion-research-documentation | 5 | 6mo | No flags | Intermediate |
| search-company-knowledge | 2 | 5mo | No flags | Beginner |
| mq | 0 | 4mo | Review | Intermediate |
Try saying
Example prompts that trigger this skill in your AI assistant.
You might also like
notion-research-documentation
openai
Research across Notion and synthesize into structured documentation; use when gathering info from multiple Notion sources to produce briefs, comparisons, or reports with citations.
search-company-knowledge
atlassian
Search across company knowledge bases (Confluence, Jira, internal docs) to find and explain internal concepts, processes, and technical details. When Claude needs to: (1) Find or search for information about systems, terminology, processes, deployment, authentication, infrastructure, architecture, or technical concepts, (2) Search internal documentation, knowledge base, company docs, or our docs, (3) Explain what something is, how it works, or look up information, or (4) Synthesize information from multiple sources. Searches in parallel and provides cited answers.
mq
muqsitnawaz
Query markdown, HTML, and PDF files with mq CLI. Triggers on: exploring doc structure, extracting sections from large .md/.html/.pdf files, 'use mq', or when reading full documents wastes tokens.
kb
compozy
Comprehensive skill for the `kb` CLI and the Karpathy Knowledge Base pattern. Covers the full KB lifecycle — topic scaffolding, multi-source ingestion (URLs, files, YouTube videos and channels, Instagram reels, bookmarks, codebases), wiki article compilation, cross-article querying with file-back, l
zotero
Agents4Academia-AI
Ingest a Zotero PDF (with annotations) into a clean source file + a consolidated annotation note
ml-paper-writing
davila7
Write publication-ready ML/AI papers for NeurIPS, ICML, ICLR, ACL, AAAI, COLM. Use when drafting papers from research repos, structuring arguments, verifying citations, or preparing camera-ready submissions. Includes LaTeX templates, reviewer guidelines, and citation verification workflows.