Skip to content

Releases: masterleopold/transmute-framework

v3.0.0 — 22nd-Pass Audit Sync: Full Template Parity

17 Mar 11:05

Choose a tag to compare

Safety-Critical Gate Fixes

  • 6V gate: Corrected to "worse-of" dual system (percentage-based AND fixability-based)
  • Post-6R FAIL routing: Corrected to "fix 6V-C manually" (human-judgment issues)
  • 6P max-cycle exhaustion: Expanded with "effectively converts to CONDITIONAL PASS" framing
  • Stage 1 gate: Reclassified — no formal PASS/FAIL gate
  • 6V/7V scenario caps: Split into distinct rows (150 for 6V full, 15 for 7V smoke)

Template Files (all synced byte-identical with canonical template)

  • CLAUDE.md: Missing Stage 3 prerequisite, expanded 6R skip condition, 6D skip condition, Stage 8 WARN correction
  • execution-guide.md: 13 differences resolved including 6P-R prerequisite precision, credential tiers format, Stage 8 batch limit removal
  • feature_scenario_generation.md: 10 updates — Step 0 filter, generalized paths, cycle detection procedure, Mermaid validation, flaky handling
  • All 6 rule templates: Cross-reference corrections, baseline rule count, EMPTY_COMPONENT comment, testing cross-reference

Skill Detailed Guides (all 23 replaced with canonical prompt files)

  • All stage prompts now match the 22nd-pass audit versions exactly
  • Stale duplicate reference files removed
  • SKILL.md reference paths updated

Agent Definitions (all 7 updated)

  • feature-tests: Narrowed to E2E-only
  • feature-reviewer: Checklists aligned (10 items, blocking/secondary)
  • feature-backend: Crash recovery, Session Language, 13 backend testing sub-rules
  • feature-frontend: 7 frontend testing pitfalls, expanded design guidelines, i18n
  • brd-writer: Fixed domain naming (3 mismatches), dynamic token budget
  • prd-writer: Dynamic token budget, BRD issues handling
  • transmute-pipeline: Stage 4 adds .claude/rules/*.md verification

v2.9.0 — 17th-Pass Audit Sync

17 Mar 11:04

Choose a tag to compare

17th-pass audit sync: gate precision, cross-template consistency, Session Language propagation, safety hardening.

Synced from canonical Transmute Framework Template (15th–17th pass audits).

v2.8.0 — 14th-Pass Audit Sync: Gate Precision & Crash Recovery

17 Mar 11:04

Choose a tag to compare

Templates

  • CLAUDE.md: "Never skip" includes 6P/6P-R; Stage 7 prerequisites tightened; 6R skip conditions add 6V FAIL handling; Stage 8 prerequisite clarifies 7D FAIL blocks; Stage 5 resumption includes 🔧 In Progress
  • execution-guide.md: 6F output adds package.json; Node.js prerequisite v20.17+; 6V routing clarified; 6R max-cycle report annotation; PRODUCTION_TEST_USERS note; 5B category prefix changed to 5B-A/5B-B/5B-C; Stage 5 completion gate expanded
  • rules-templates: REST/GraphQL choice comment; CLAUDE.md Part 2 sync requirement; explicit default timeout values

Skills (19 updated)

  • audit-a11y, audit-security, implement, feedback, redesign, brd, docs, prd, scaffold, harden, audit-completeness, maintain, optimize, smoke, remediate, tech-stack, validate-specs, verify, polish
  • Key changes: BRD file grep-based search, crash recovery protocol, retry tracking, Bun v1.2+ support, interrupted 6P-R detection

v2.7.0 — 12th-Pass Audit Sync

17 Mar 11:04

Choose a tag to compare

12th-pass audit sync: 7V CONDITIONAL PASS gate, safety rule refinements, prerequisite alignment.

Synced from canonical Transmute Framework Template (11th–12th pass audits).

v2.6.0 — 10th-Pass Audit Sync: Full Reference Parity

17 Mar 11:04

Choose a tag to compare

Templates

  • CLAUDE.md: 5B restored to "Never skip" list; Stage 6 ordering "(mandatory — not merely recommended)"; notation explanation expanded; 6D prerequisite restored; Stages 8+9 conflict list restored; candidate staging 30-limit
  • execution-guide.md: "Recommended" → mandatory ordering; inline ordering notes restored; 6H detailed goal; 6P-R max-cycle exception; 5B unskippable
  • rules-templates: 4-step instruction with HTML comment guidance; concrete env var example; conditional OMIT for schemaless backends; inline SVG qualifier

Skills (Full Reference Parity)

  • All 23 skills: Reference files now contain complete template prompt content
  • 10 new reference files created: validate-specs, audit-security, audit-a11y, optimize, seed-data, harden, docs, prelaunch, feedback, maintain
  • 13 existing reference files expanded: tech-stack (+780 lines), verify (+727 lines), implement (+468 lines), and more

Terminology

  • "QA & Hardening" → "Quality Assurance" across pipeline agent, README, and all references

v2.5.0 — 7th-Pass Audit Sync: Gate Edge Cases & Template Precision

17 Mar 11:04

Choose a tag to compare

Templates

  • CLAUDE.md: Recovery procedure cross-references expanded; staleness policy clarified; added style commit type
  • execution-guide.md: Assumption volume formula; 6V/6R B-category fixability criteria; CONDITIONAL PASS evaluation order; flaky scenario definition
  • feature_scenario_generation.md: Step 5 restructured; estimation time breakdown; personas scoring aligned
  • _rules-candidates.md: Restored missing Date Added field (breaking fix — staleness policy requires this)
  • rules-templates: Frontend restored "Additionally:" prefix; Testing comment metadata aligned

Skills

  • verify, remediate, maintain — B-category fixability criteria, staleness review updates

v2.4.0 — Comprehensive Template Sync & Agent Hardening

17 Mar 11:04

Choose a tag to compare

Templates

  • CLAUDE.md: Structural overhaul — concise Safety-Critical Rules + Cross-References table pattern; inline style exception; third-party any type wrapping; traceability exemptions
  • execution-guide.md: 30+ changes — scaffold coverage, 5B gate tightening, 6R cycle reset, assumption review timing, credential canonical source, recovery procedures
  • feature_scenario_generation.md: Unresolvable cycle fallback, scenario cap override, _progress.md fallback
  • _rules-candidates.md: Staleness policy fix (OR condition), overflow handling at 30 candidates
  • rules-templates (all 6): grep pattern fix, TODO format, sanitization guidance, auth checks, test count preservation

Agents (all 7 updated)

  • transmute-pipeline: Full Stage 7 prerequisite chain, 5B FAIL-RETRY thresholds, per-feature escalation tracking
  • brd-writer/prd-writer: Known Failure Patterns, Language Rule, Deduplication Rule
  • feature-backend/frontend/tests: Known Failure Patterns, Cross-Feature Integration

Skills

  • scaffold, implement, audit-completeness, audit-security, verify, remediate, smoke, docs, seed-data

Hooks

  • Created check-prerequisites.sh — was referenced by hooks.json but missing (blocking bug fix)

v2.3.0 — Gate Precision & Terminology Alignment

17 Mar 11:04

Choose a tag to compare

Gate Updates

  • 5B: Priority-order evaluation, expanded CONDITIONAL PASS paths (a/b/c), per-feature run state
  • 6R: 6V-A/6V-B/6V-C terminology, auto-escalation after 3 cycles, max 2 outer 6V→6R cycles
  • 6P: Category E discretionary skip conditions
  • 6P-R: Severity classification (Critical/Major/Minor)
  • 3: CONDITIONAL PASS requires Part 2 + _progress.md
  • 2B: Coverage definition disambiguation, FAIL edge case examples

Skill Updates

  • Scaffold, 6G, Remediate, Stage 9, Stage 0, User guide, Feedback, Progress tracking
  • Stage 8 prerequisite: 7D PASS/WARN check added

v2.2.0 — Comprehensive Template Sync (2-Pass Deep Audit)

17 Mar 11:04

Choose a tag to compare

Rules Templates (all 6 updated)

  • Added Source: and Confidence: metadata to all <!-- TODO --> comments
  • Added post-render verification grep command to all template headers
  • Backend, Data Model, Auth, Testing, Frontend, API Contracts — all expanded

New Template File

  • _rules-candidates.md — staging area for rule candidates from Stages 5B and 6R

CLAUDE.md Template Updates

  • Pre-6V Setup, Stage 0/3 expansions, 6A/6G scope, 6V modes, 5B/6V gate details
  • Test Count Preservation rule added to Part 1

Execution Guide Updates

  • 6G naming, 6R terminology, 6V/6R category prefix alignment
  • Pre-6V Setup reorganized, Deployment Failure Recovery section added

Feature Scenario Generation Updates

  • Implicit dependency algorithm, cycle detection, negative variant counting
  • User impact scoring algorithm, flaky scenario handling

v2.1.0 — 6R Cycle Semantics

17 Mar 11:03

Choose a tag to compare

6R Cycle Semantics

  • Clarified "Max 3 completed loops" → "Max 3 internal fix-verify cycles within a single 6R run"
  • Added explicit FAIL recovery guidance

6V Routing

  • PASS → skip 6R, CONDITIONAL PASS with 6V-A/B → run 6R, CONDITIONAL PASS with only 6V-C → skip 6R

Other

  • 6V- prefix consistency across all skills, agents, and templates
  • New operator checkpoint between Stages 1 and 2B for assumption review
  • Stage skill updates: 5B, 6P, 6P-R, 6R
  • Template updates: CLAUDE.md, execution-guide.md, API contracts