Gabe Scope Pivot

Gabe Scope Pivot — direction-change rewrite

Usage: /gabe-scope-pivot <description>

Gabe execution contract (E1–E7)

This skill runs under the suite execution contract — E1 EVIDENCE · E2 RUN-BEFORE-✅ · E3 NO SILENT DOWNGRADE · E4 REUSE FIRST · E5 STATE SYNC · E6 MISSING ANCHOR = STOP · E7 REPORT WHERE — floors, not ceilings; a skill's own gate may be stricter, never looser. Full text: ../gabe-docs/references/execution-contract.md (if that file is missing, E6 applies — STOP).

Handles direction-shifting scope changes. Unlike addition, a pivot restructures the premise — downstream SCs, REQs, phases, and refs all need re-evaluation. The old SCOPE is archived (never deleted); a new v{N+1} is derived.

Invariants:

• Old SCOPE.md v{N} + scope-references.yaml v{N} archived to .kdbp/archive/vN/ (a legacy standalone .kdbp/ROADMAP.md, if the project still has one from before A2, is archived alongside it)
• New SCOPE.md scope_version: bumps to {N+1}; last_scope_event: updates
• New SCOPE.md phases_version: resets to 1 (new scope = new phase-arc history)
• Pivot rationale recorded in Change Log with reference to archived v{N}
• Research directory re-archived if regenerating (optional prompt)

Rendering note. Output templates in this spec wrapped in bare triple-backtick fences are spec-meta delimiters — render their contents as plain markdown at runtime. Tagged fences (yaml, diff) stay fenced. See gabe-docs/SKILL.md § "Runtime output rendering convention".

Procedure

Step 1: Pre-flight

Same as /gabe-scope-change — SCOPE.md must exist, no active session.

Announce the pivot intent:

PIVOT WARNING

This will archive SCOPE.md v{N} (created {date}) to .kdbp/archive/v{N}/
(a legacy standalone ROADMAP.md, if present, is archived alongside it).
A new v{N+1} will be derived from the changed premise.

Pivot reason: {description}
Trigger rule: {from classifier}

This is a disruptive change. Downstream SCs, REQs, and phases will
be re-evaluated. Open questions from v{N} transfer to v{N+1}.

Type `pivot to v{N+1}` to confirm:

Typed confirm (exact match, case-insensitive). Any other response → abort.

Step 2: Archive v{N}

mkdir -p .kdbp/archive/v{N}/
cp .kdbp/SCOPE.md .kdbp/archive/v{N}/SCOPE.md
cp .kdbp/scope-references.yaml .kdbp/archive/v{N}/scope-references.yaml
# Legacy projects (pre-A2) may still carry a standalone ROADMAP.md — archive it too if present.
if [ -f .kdbp/ROADMAP.md ]; then
  cp .kdbp/ROADMAP.md .kdbp/archive/v{N}/ROADMAP.md
fi
# Research: archive the live tree; new pivot session will regenerate
if [ -d .kdbp/research ]; then
  mv .kdbp/research .kdbp/archive/v{N}/research
fi

Never delete. Archives are permanent audit trail.

Step 3: Re-derive v{N+1}

Pivots require re-reasoning, not just mechanical transformation. Invoke the same step sequence as /gabe-scope but with prior state loaded:

(a) Seed session. Create fresh scope-session.json with:

• current_step: step-1-intake
• Context carried forward from v{N}: intake_summary + research_summary (as starting prior, not gospel)
• Pivot rationale + description stashed in session for use at Step 3 (Problem + Vision)

(b) Run interview. Step 1 re-runs but pre-answers are seeded from v{N} intake. Opus intake-quality-evaluator re-checks each against the pivot description — answers inconsistent with the new direction get re-asked.

(c) Research. Step 2 research can be skipped (user prompt: "Re-run research or reuse v{N} research?"). If reuse, the archived research tree is un-archived for the new session.

(d) Drafts and finalize. Steps 3–8 run as normal. Opus is explicitly told via the reasoning context: "This is a pivot of v{N}; the new primary direction is {description}. Inconsistencies with v{N} are expected and intended."

Step 4: Pivot-specific Change Log entry

The init Change Log row in v{N+1} SCOPE.md is replaced with a pivot row:

## 15. Change Log

| Date | Type | Summary |
|---|---|---|
| 2026-04-22 | pivot | Pivoted from v1 (archived at .kdbp/archive/v1/). Reason: {description}. Trigger: {trigger_rule}. Prior version's Open Questions migrated. |

Step 5: CHANGES.jsonl entry

{"ts":"2026-04-22T10:00:00Z","event":"scope_pivot","from_version":1,"to_version":2,"trigger_rule":"primary_user","archive_path":".kdbp/archive/v1/","rationale":"{description}"}

Step 6: Migrate Open Questions

Open Questions from v{N} that weren't resolved carry to v{N+1} §14 Open Questions with a [migrated from v{N}] tag. User prompted per-OQ: keep / resolve-now / drop-as-obsolete.

Step 7: Finalize

Same as /gabe-scope Step 8 — write files, archive research (of the new session), tombstone session.json, git-commit prompt.

Flags

• --skip-research-prompt — don't ask about research reuse; always re-run.
• --reuse-intake — skip Step 2(b) interview re-check; accept all v{N} intake answers as-is.

Edge cases

Pivot without confirmation. Never proceed without typed confirm. No --yes flag.

Pivoting an already-pivoted scope. v{N} = 3 → archive to .kdbp/archive/v3/, new version 4. No cap on pivot depth.

Archive directory already exists. Abort with: "Archive .kdbp/archive/v{N}/ already exists. Prior pivot not completed cleanly?" Manual cleanup required before proceeding.

Pivot description empty. Prompt; exit on second empty attempt.

User aborts during re-interview. Session.json persists in the in-progress state. On next invocation: "Pivot in progress at Step {N}. Resume or abort?" If abort, the archive is preserved (v{N} files stay in archive); active SCOPE.md gets restored from archive; no partial v{N+1} emitted.

Integration

Called by /gabe-scope-change (routed) or directly (not recommended without classifier rationale).

/gabe-plan warned on next invocation if SCOPE.md phases_version changed: "Phase arc regenerated. Re-read phase context?"

Command version

v1.0.