Graduate a concept-forge artifact into the vault as a network of notes โ spine concept + card-grade satellites + person-note updates + sources propagation + provenance split + sibling-cluster auto-offer. Use when the user says to integrate a concept-forge artifact, graduate a forge, or close out a forging session.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
A direct command skips the review prompt. Inspect the source before running it.
Graduate a concept-forge artifact into the vault as a network of notes โ spine concept + card-grade satellites + person-note updates + sources propagation + provenance split + sibling-cluster auto-offer. Use when the user says to integrate a concept-forge artifact, graduate a forge, or close out a forging session.
argument-hint
[forge-slug or path to Workbench/concept-forge/[slug].md]
Integrate Concept-Forge: $ARGUMENTS
Graduate a concept-forge artifact at Workbench/concept-forge/[slug].md into a network of permanent vault notes โ spine concept + card-grade satellites + person-note touches + sources propagation + provenance split + auto-offered chain to /integrate-concept-cluster if a sibling workbench cluster exists.
This is the forge-specific integration skill. For non-forge workbench items use /integrate-workbench (lightweight default). For workbench clusters use /integrate-concept-cluster (this skill auto-offers the chain at Step 11).
Reference rules and global constraints
Load: concept-craft (concept card schema, anti-sycophancy/refinement-addiction guardrails, "tool is not truth"), concept-forge-artifact-format (specification of the workbench file format being parsed โ frontmatter, body sections, templates), workbench (graduation conventions), operating-principles (proactive linking checklist), vault-conventions (atomic-note discipline + collection-note exception, frontmatter conventions, 7-folder structure).
Global constraints:
Plan-before-execute โ Never modify vault state until user confirms each phase. Show the plan, wait for approval, then execute.
Per-item user gating โ Heuristics propose; user disposes. Always.
Tool is not truth โ When the schema or heuristic doesn't fit, the schema is wrong, not the content. Honor user override always.
Self-similarity check โ The heuristics introduce structure. Don't let structure force decomposition where coherence wants preservation.
Dynamic Context
Forge artifact preview
!obsidian read path="Workbench/concept-forge/$ARGUMENTS.md" vault="{{VAULT_NAME}}" 2>/dev/null | head -60 || echo "(forge artifact not at Workbench/concept-forge/$ARGUMENTS.md โ try resolving slug in Step 1)"
Sibling cluster preview
!ls -la "Workbench/$ARGUMENTS/" 2>/dev/null || echo "(no direct sibling cluster at Workbench/$ARGUMENTS/)"
Step 1: Locate and validate the forge artifact
Parse $ARGUMENTS:
If a path: read directly via obsidian read path=....
If a slug: resolve to Workbench/concept-forge/[slug].md.
If ambiguous or empty: search via obsidian search query="path:Workbench/concept-forge/ $ARGUMENTS" and ask which to integrate.
Read the full file. Verify frontmatter status: ready. If status: drafting or paused, confirm: "This forge is [status] โ proceed with integration?"
Step 2: Detect binding-argument scaffold
Aggressive decomposition requires a binding summary in the spine that holds structure together after extraction. Scan ## Concept for a section named like:
"The Argument, Condensed"
"The Core Argument"
"Summary"
"The Move(s)"
Or content that functions as a numbered binding-summary at the start of ## Concept
If absent, warn:
No binding-argument scaffold detected. Aggressive extraction risks fragmenting the concept into a list of pointers without binding structure. Options:
(a) Pause; write a binding summary first (recommended) โ exit skill, user edits forge, re-runs.
(b) Conservative extraction โ only the 1-2 strongest satellites.
(c) Proceed aggressive anyway (not recommended).
Wait for user choice. If (a): exit. If (b): note conservative mode for Step 3. If (c): proceed.
Step 3: Decomposition proposal โ PLAN, do not execute
Apply the card-grade heuristic from references/card-grade-heuristic.md to each subsection of ## Concept. Read that reference file before scoring. It contains the 4-criterion rubric, the threshold rule (2+ checks โ satellite candidate), the destination-hint logic (evergreen vs scoped), and worked-example calibration.
For each candidate: compute the score, classify destination (evergreen / scoped / mixed), and group by kind:
Vocabulary moves
Strategic frames
Thought experiments / lenses
Deployable principles
Present grouped proposal with per-candidate score visible:
User accepts/skips by group with granular per-item override. Do not proceed until user confirms. If conservative mode (Step 2 fallback), propose only the top 1-2 highest-scoring candidates.
Step 4: Spine-preservation check
Build the post-extraction spine outline from accepted satellites in Step 3. Show the user the spine as it will look after extraction:
Spine after extraction:
[Forge Title]
Core Insight โ preserved
The Argument, Condensed โ preserved with cross-refs to [[Satellite 1]], [[Satellite 2]] ...
Purpose Served / Anti-Scope / Nebulosity Type โ preserved
Counterarguments & Responses โ preserved
Satellites โ list of wikilinks
Concur, or pull anything back into the spine?
User can pull items back into the spine (extraction skipped for those items; content stays inline). Do not proceed until user confirms.
Step 5: Person-note phase
Aggregate verbatim quotes + contributions per person across ## Concept and ## Session Log. Use grep / obsidian search across the forge artifact to find every quote attribution and ratification marker.
For each person, present a per-person gate:
[Person Name] has N contributions to this forge:
[contribution one] (date / context)
[contribution two] (date / context)
...
Add all N to References/[Person Name].md? Some? Skip?
If person note doesn't exist, propose creation using the person template (search Templates/ for the canonical person template). User accepts/amends per person. Do not proceed until user confirms each person.
Step 6: Sources phase
Count sources in ## Sources & References. Group by thematic clusters (subsection headers in that section).
Heuristic: 8+ total sources OR sources span 3+ thematic clusters โ propose collection note at Notes/[Topic] Reading List.md (intentional anti-atomic exception per vault-conventions). Else inline in spine's Sources section.
For collection-note proposal, draft a one-line description and propose category note. User confirms/amends. Do not proceed until user confirms.
Step 7: Auto-split provenance (no user gate)
Move ## Session Log content from forge artifact to a new References/Forge History โ [forge-name].md. If ## Spawn Candidates exists in the forge artifact (any status), include it in the same Forge History split โ same provenance shape. Spawn decisions are part of the forge's design rationale and belong in the historical record, not in active retrieval. Active spawned forges retain their links via ## Cross-links (which propagate to the spine concept note via Step 10).
The spine note (built in Step 8) keeps a footer pointer: **Forge history**: [[Forge History โ [name]]]. This step has no user gate โ it's a settled default per design.
Step 8: Build the spine concept note
From the (now-thinner) ## Concept block โ the spine after Step 4 confirmation โ draft the spine note. Reference references/card-schema.md for the Chapman card schema and per-field guidance, especially the Spine note structure section at the end.
Use Edit tool for safe placement (never obsidian create overwrite). Apply the full proactive linking checklist from operating-principles.md:
Set categories (typically [[Concepts]] or [[Evergreen]]; ask user if ambiguous).
Set areas (inherit from forge frontmatter).
Set status: active.
Verify tags; ensure created is set.
Search vault for related notes (QMD vec query with intent + obsidian search for keywords); add wikilinks where meaningful.
Verify category note exists; offer to create if missing.
Place at Notes/[Forge Title].md. Add the forge-history footer pointer.
Step 9: Build satellite notes
For each accepted satellite from Step 3, draft a Chapman card note. Read references/card-schema.md first for per-field guidance; that file documents how to source each field from the forge subsection.
For each satellite:
Draft using the Chapman card schema (Core Insight; Purpose Served; Anti-Scope; Nebulosity Type; The Insight Unpacked; Bounds & Edges; Analogies & Connections; Prior Art; Counterarguments & Responses; Revision Conditions; Open Questions). Fields may be blank where the forge content didn't develop them.
Source content from the corresponding subsection of the forge's ## Concept. Pull verbatim quotes; cite sources from ## Sources & References where they're cited there.
Apply proactive linking checklist (categories per Step 3 destination; areas inherited; related notes searched).
Route to destination โ evergreen โ Notes/[Satellite Title].md with categories: ["[[Concepts]]"] or ["[[Evergreen]]"]; scoped โ Notes/[Satellite Title].md with project link in body.
Cross-link: each satellite links back to the spine; relevant sibling satellites link to each other.
Step 10: Cross-link propagation
Search the vault for related concept notes that should reference the new vocabulary.
Use QMD: mcp__qmd__query with:
searches: [{type:'vec', query:'[satellite name] OR [spine concept name]'}]
intent: "finding existing concept notes that should reference [new vocabulary] from this forge"
For accepted updates, propose the specific edit (typically a wikilink + one-sentence context). User accepts/skips per note. Do not proceed until user confirms each.
Step 11: Sibling-cluster auto-offer
Check for adjacent workbench cluster:
Workbench/[Forge Title]/ โ direct subdirectory
Workbench/[adjacent name]/ โ semantically related workbench subdirectory referenced in the forge body
If detected:
Cluster detected at [path] (N files). Run /integrate-concept-cluster now? It will receive the satellite-decision manifest from this run as context.
(a) Yes, run now โ chain in same session.
(b) Later โ I'll run /integrate-concept-cluster standalone.
If (a): invoke /integrate-concept-cluster with satellite-decision manifest passed through (which content was absorbed into which satellite, so cluster triage can reference it). If (b) or no cluster detected: skip.
Step 12: Leave the forge artifact in place
Do NOT move, archive, or delete the forge artifact. The artifact remains at Workbench/concept-forge/[slug].md after integration.
User direction: artifact still serves as design rationale during sibling skill development; user may process it later (potentially via this skill on itself as a self-test once both integration skills are stable).
Surface a one-line note: "Forge artifact left in place at Workbench/concept-forge/[slug].md โ process later when ready."
Use specific file adds, not git add -A. Forge artifact stays in workbench (untouched), so it's not in the commit.
Edge Cases
Forge status: drafting or paused โ Confirm before proceeding; the artifact may not be ready.
No satellite candidates (all subsections score <2/4) โ Skip Steps 3-4. Single-concept integration: just the spine note. Run Steps 5-13.
No binding-argument scaffold + user picks (b) โ Conservative extraction: top 1-2 strongest candidates only.
Cluster-only graduation needed โ User wants to integrate cluster but not concept yet. Direct them to /integrate-concept-cluster standalone; exit this skill.
User aborts mid-flow โ No commits made; nothing on disk has changed. Confirm: "Aborting. Forge artifact and vault unchanged. Re-run when ready."
Person already has a note + new contributions โ Read existing note; append new contributions section dated by forge graduation; preserve existing content.
Satellite name collides with existing note โ Prompt: "Existing [[Note Name]] found. Merge into existing, rename satellite, or skip?"
Anti-pattern reminders
Never silently mutate vault state. Every phase shows the plan; user confirms; then execute.
Never collapse user gating into batch-of-one. Group items, but always allow granular per-item override.
Never force decomposition where coherence wants preservation. If the user says "leave it in spine," honor it absolutely.
Never invent verbatim quotes. Person-note contributions cite the artifact's actual quotes โ copy verbatim, attribute by date.
Honor "tool is not truth" (concept-craft). When the heuristic or schema doesn't fit, the heuristic is wrong, not the content.
