mind-merge
Adjudicate long or branching design conversations by maintaining a local graph of open Questions, accepted Decisions, Evidence, dependencies, provenance, and…
Skill by ElmoJones3
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/elmojones3-mind-merge
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: design, frontend, productivity
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
- Runs as
- Node.js package
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/mind-merge
What's inside
- Skill name
mind-merge- Description
- Adjudicate long or branching design conversations by maintaining a local graph of open Questions, accepted Decisions, Evidence, dependencies, provenance, and Phase changes in a durable HTML worksheet. Use only when explicitly requested.
- License
- MIT
- Instructions
- 64 lines in SKILL.md · view on GitHub
Before installing
- Ships executable scripts
Files (35)
- .gitignore
- LICENSE.matt-pocock
- SEMANTICS.md
- SKILL.md
- THIRD_PARTY_NOTICES.md
- openai.yaml · agents/
- mind-merge.mermaid.html · assets/
- mind-merge.rf.html · assets/
- package-lock.json
- package.json
- adjudication.md · references/
- state-model.md · references/
- state.schema.json · references/
- worksheet.md · references/
- build-assets.mjs · scripts/
- capture-phase.mjs · scripts/
- init-session.mjs · scripts/
- cli.mjs · scripts/lib/
- document.mjs · scripts/lib/
- session.mjs · scripts/lib/
- yaml.mjs · scripts/lib/
- render-session.mjs · scripts/
- validate-state.mjs · scripts/
- mermaid-app.mjs · template-src/
- rf-app.jsx · template-src/
- shared.mjs · template-src/
- shell.html · template-src/
- styles.css · template-src/
- yaml-vendor-entry.mjs · template-src/
- browser-smoke.browser.mjs · tests/
- cli.test.mjs · tests/
- docs.test.mjs · tests/
- schema.test.mjs · tests/
- session.test.mjs · tests/
- template.test.mjs · tests/
SKILL.md (first lines)
# Adjudicate shared understanding Act as the Adjudicator. Own the mechanics of exposing Questions, recognizing acceptance, capturing each Phase, deriving the frontier, and posing the next meaningful Question. Do not take ownership of the user's semantics, philosophy, authority, or priorities. Read [SEMANTICS.md](SEMANTICS.md) and [references/adjudication.md](references/adjudication.md) before creating, recovering, or changing session state. Read [references/state-model.md](references/state-model.md) when interpreting the graph or handling validation failure. Read [references/worksheet.md](references/worksheet.md) when initializing, capturing, validating, rendering, or rebuilding artifacts. ## Establish durable state Search for an existing `state.json`, `state.yaml`, or Mind Merge worksheet before relying on conversation memory. - If state exists, validate and render it. Treat it as the authority for settled Decisions and provenance. - If no state exists, initialize from the current conversation. Backfill only explicit acceptance; record facts as Evidence and uncertainty as Questions. - Use the repository's design-artifact convention. If it has none, use `docs/mind-merge/<subject-slug>/`. - Capture time, branch, and Git user explicitly at the boundary. Do not read ambient time, Git, or files inside the pure state operations. Canonical state is JSON or YAML. The worksheet is a generated projection, not an authority. ## Run the adjudication loop 1. Validate the state and derive Question states, active edge lineage, dependencies, and the frontier. 2. Open any Questions exposed by the latest answer or Evidence. 3. Resolve facts through inspection and derive mechanical consequences from accepted Decisions and repository laws. 4. Posit one frontier Question that genuinely requires user judgment and most reduces what remains. 5. Ask it plainly. Use a counterexample, boundary case, or identity test only when it can change a live distinction. 6. Distinguish exploration from acceptance. Never convert tentative discussion into a Decision. 7. Capture the Phase after an accepted answer or after exploration materially changes the graph. 8. Validate and regenerate the worksheet. 9. Pose the next nominated Question. Do not ask the user to choose file paths, IDs, timestamps, edge mechanics, renderer internals, or consequences already fixed by accepted laws. Those are Adjudicator work. ## Capture provenance - Keep Questions, Decisions, and Evidence separate. - Connect Decisions to Questions only through independently identified Provenance Edges. - Use only `raises`, `resolves`, `supersedes`, `reopens`, and `retracts` as state-bearing edge types. - Point every later edge on a Question to its current edge with `supersedesEdgeId`. - Require a new accepted Decision for every retraction. - Model Question prerequisites with immutable Question Dependencies; their relationship is `requires`. - Put subject-specific explanation in open metadata. Metadata never changes derived state. - Never store Question state, the active graph, or the frontier as mutable fields. An unresolved disagreement remains Evidence plus an open Question. Closing a bounded pass may leave it visible; closure does not invent acceptance. ## Capture before continuing Use the supplied commands rather than hand-editing canonical state: ```bash node <skill-directory>/scripts/init-session.mjs --input init.json --state state.json node <skill-directory>/scripts/capture-phase.mjs --state state.json --input phase.json node <skill-directory>/scripts/validate-state.mjs --state state.json node <skill-directory>/scripts/render-session.mjs --state state.json --renderer mermaid --output mind-merge.html ``` Use the Mermaid renderer for the compact default. Use `--renderer rf` when pan, zoom, minimap navigation, or entity inspection materially helps. Both templates are offline and project the same state. Visual interaction owns layout and viewport only. The capture command appends all accepted changes or refuses them without altering the state file. Never ask the next Question until the Phase that nominates it is durable. ## Finish the pass Close when the bounded frontier is exhausted or the user chooses to stop. Preserve unresolved Questions. Do not declare shared understanding complete or produce the subject skill's downstream artifact until the user confirms the merge.
From ElmoJones3/opinionated, MIT.
More from ElmoJones3/opinionated
- 0.592score
domain-modeling
ElmoJones3
Make business objects own their behavior, construction, invariants, and consequences.
SkillAFreshness: A★ 1Claude Code +9 - 0.549score
principle-bound-telemetry
ElmoJones3
Bound collection cost and protect business outcomes when configuring or reviewing telemetry fields, cardinality, queues, exporters, sampling, loss reporting…
SkillBFreshness: B★ 1Claude Code +9 - 0.549score
principle-interpret-observations
ElmoJones3
Preserve evidence meaning when receiving observations, selecting current execution evidence, or calculating operational reports, rates, and coverage.
SkillBFreshness: B★ 1Claude Code +9 - 0.549score
principle-always-comment-code
ElmoJones3
Preserve the reason code exists beside every source file, named declaration, and member.
SkillBFreshness: B★ 1Claude Code +9 - 0.549score
principle-code-first-documentation
ElmoJones3
Treat generated OpenAPI and its documentation UI as part of every HTTP boundary.
SkillBFreshness: B★ 1Claude Code +9 - 0.549score
principle-observe-operations
ElmoJones3
Establish operation ownership and one coherent completion observation when adding or reviewing instrumentation across logs, traces, or metrics.
SkillBFreshness: B★ 1Claude Code +9 - 0.549score
principle-observe-progress
ElmoJones3
Define useful advancement, legitimate waits, and fresh live evidence when instrumenting or reviewing streams, long-running operations, or stall and spin…
SkillBFreshness: B★ 1Claude Code +9 - 0.549score
principle-prefer-pure-functional-patterns
ElmoJones3
Prefer pure calculations and explicit effects when creating, changing, reviewing, or diagnosing transformations, validators, reducers, pipelines, or state…
SkillBFreshness: B★ 1Claude Code +9 - 0.549score
semantic-snapshot
ElmoJones3
Review the recent conversation for settled terminology and decisions, then update semantic files and missing ADRs.
SkillBFreshness: B★ 1Claude Code +9 - 0.549score
semantic-mapping
ElmoJones3
Keep project terms and definitions consistent in SEMANTICS.md and SEMANTIC-MAP.md.
SkillBFreshness: B★ 1Claude Code +9 - 0.544score
principle-sql-preserve-query-cardinality
ElmoJones3
Preserve root selection, child completeness, ordering and bounded query work when implementing or reviewing SQL aggregate reads, filters, joins or pagination.
SkillBFreshness: B★ 1Claude Code +9 - 0.544score
stay-in-your-lane
ElmoJones3
Keep agents inside user-defined scope and authority.
SkillBFreshness: B★ 1Claude Code +9
Signals
- Stars
- 1
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Aug 22, 2026 (2 months ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/ElmoJones3/opinionated
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from ElmoJones3
- 0.592score
domain-modeling
ElmoJones3
Make business objects own their behavior, construction, invariants, and consequences.
SkillAFreshness: A★ 1Claude Code +9 - 0.549score
principle-observe-progress
ElmoJones3
Define useful advancement, legitimate waits, and fresh live evidence when instrumenting or reviewing streams, long-running operations, or stall and spin…
SkillBFreshness: B★ 1Claude Code +9 - 0.549score
principle-always-comment-code
ElmoJones3
Preserve the reason code exists beside every source file, named declaration, and member.
SkillBFreshness: B★ 1Claude Code +9 - 0.549score
principle-code-first-documentation
ElmoJones3
Treat generated OpenAPI and its documentation UI as part of every HTTP boundary.
SkillBFreshness: B★ 1Claude Code +9 - 0.549score
principle-observe-operations
ElmoJones3
Establish operation ownership and one coherent completion observation when adding or reviewing instrumentation across logs, traces, or metrics.
SkillBFreshness: B★ 1Claude Code +9 - 0.549score
principle-prefer-pure-functional-patterns
ElmoJones3
Prefer pure calculations and explicit effects when creating, changing, reviewing, or diagnosing transformations, validators, reducers, pipelines, or state…
SkillBFreshness: B★ 1Claude Code +9
Cite this page
mind-merge skill. Agent Mart. https://agentmart.dev/skills/elmojones3-mind-merge (last updated 2026-10-24)