TS — TFW-55 / Phase A.2: North Star Values and Consumer Integrity¶

Date: 2026-08-26 Author: Coordinator Status: ✅ TS_APPROVED — owner 2026-08-26; execute through a separate /tfw-handoff task Parent HL: HL-TFW-55 Phase HL: Phase A.2 derivation Pre-A7 contract baseline: e3e7b33; Amendment A7 is owner-approved and applied at the new narrow re-freeze baseline Approval gate: passed — A7, the Phase HL, this TS, both bounded retirements, the full consumer-integrity scope, and the 4,200-word ceiling were explicitly approved 2026-08-26

1. Objective¶

Complete the good current Project North Star without rolling it back: explicitly dispose all eight TFW-25 values and the four TFW-32 team-centric Success Criteria, restore the mandatory current values that were silently lost, and repair every active normative/current consumer so Project Values priority 0/1 citations resolve to the meaning they claim. Preserve the old Phase A APPROVE as an honest trace of work accepted against an incomplete scope, preserve the shared dirty workspace exactly, keep the essay at no more than 4,200 descriptive whitespace-delimited words without treating that maximum as a target, and do not begin the BoK.

2. Scope¶

In Scope¶

bounded editorial integration in the current .tfw/README.md, preserving its essay form and owner brand-image insertion;
a complete value/Success-Criteria disposition ledger using EXPLICIT RESTORE, SEMANTIC MERGE, or INTENTIONAL RETIRE with reason and canonical target;
explicit canonical places for Candor Over Flattery, Structural Enforcement, Naming Creates Behavior, Portability, and bounded team-centric Success Criteria;
semantic preservation of selected durable Trace/continuity without restoring code-disposability or deterministic-regeneration claims;
current PV priority 0/1 targets and semantics in glossary, conventions, the HL template, plan workflow, review workflow, and review verification template;
exact synchronization of installed Claude and Antigravity plan/review workflow copies;
correction of the current TFW-60 master HL Project North Star header and free §7.2 citations only;
new Phase A.2 ONB, EV, and RF with exact dirty-baseline and final-preservation evidence;
a separate formal Reviewer after execution, with at most three review returns before owner escalation;
post-APPROVE /tfw-docs correction of current knowledge/debt indexes as a later Coordinator stage.

Out of Scope¶

wholesale rollback or rewrite of the current North Star essay; verbatim restoration of the old Values section;
any change to the old Phase A/B/B.2 RF, REVIEW, EV, ONB, stage files, TFW-25/TFW-32 artifacts, changelog history, or Git history;
changes to root README prose/localizations, Editions, product/runtime code, adapters other than exact installed workflow copies, brand assets, docs site, or TFW-60 frozen claims/parallel Phase A draft;
BoK, methodical guide, course, book, certification, claims registry, or governance subsystem;
new /tfw-research work absent a new material uncertainty approved by the owner;
/tfw-docs before formal REVIEW APPROVE, role collapse between Executor and Reviewer, or execution outside the owner-approved A7/HL/TS scope.

3. Principles Check¶

#	Principle (master HL §7)	Enforced by	Gate
P1	One North Star, several depths	AC-3, AC-5	One essay owns purpose/principles/non-goals; consumers point to it rather than creating another authority
P2	Philosophy before machinery	AC-3, AC-4	Editorial review preserves the problem-led essay and integrates values as philosophy, not a mechanics inventory
P3	Human purpose remains human	AC-3, AC-4	Human judgment/acceptance remains explicit; the no-manual-editing absolute is retired
P4	Trace, not transcript	AC-2, AC-4	`Traces Over Code` survives only as selected durable continuity, not transcript, code disposal, or identical reproduction
P5	Self-awareness must be operational	N/A	Phase A.2 does not change the already-reviewed operational self-awareness section; scope and hash checks prevent accidental drift
P6	Function before compression	AC-2, AC-3	No approved value or outcome disappears to meet the ceiling; an overrun stops for amendment
P7	Provenance before polish	AC-1, AC-2, AC-8	Exact baselines, dispositions, old verdict, active/history boundary, and foreign state remain visible
P8	Experience reveals the system	AC-4	Bounded team-resumption and knowledge-compounding outcomes remain observable success tests
P9	Refutation before canonicalization	AC-2, AC-4	Every source item receives a reasoned disposition; contradictions drive explicit retirement rather than silent omission

4. Affected Files¶

Executor production and trace scope¶

File	Action	Description
`.tfw/README.md`	MODIFY	Coherent value/Success-Criteria integration within the existing North Star essay and 4,200-word maximum; no filler or expansion target
`.tfw/glossary.md`	MODIFY	Point PV priority 1 to actual current methodology-value clauses and define semantic scan/verification expectations
`.tfw/conventions.md`	MODIFY	Change Knowledge Citations verification from existence-only to existence plus semantic relevance, especially priorities 0/1
`.tfw/templates/HL.md`	MODIFY	Replace stale North Star example and require a full priority 0–4 scan with semantically specific citations
`.tfw/workflows/plan.md`	MODIFY	Make Scan Project Values include priorities 0–4 in full and require meaning-level priority 0/1 citation
`.tfw/workflows/review.md`	MODIFY	Make Reviewer verification include a full PV scan and semantic comparison to claimed priority 0/1 clauses
`.tfw/templates/review/verify.md`	MODIFY	Add semantic relevance and priority 0/1 clause checks; a resolving but irrelevant citation is a discrepancy
`.claude/commands/tfw-plan.md`	MODIFY	Exact synchronized copy of canonical `plan.md`
`.agent/workflows/tfw-plan.md`	MODIFY	Exact synchronized copy of canonical `plan.md`
`.claude/commands/tfw-review.md`	MODIFY	Exact synchronized copy of canonical `review.md`
`.agent/workflows/tfw-review.md`	MODIFY	Exact synchronized copy of canonical `review.md`
`tasks/TFW-60__conflict_resistant_shared_workspace/HL-TFW-60__conflict_resistant_shared_workspace.md`	MODIFY	Correct current header and free §7.2 citations; preserve every frozen claim and concurrent Phase A work
`phase-a/ONB__phase-a2__north_star_values_and_consumer_integrity.md`	CREATE	Executor source reading, exact baseline manifest, scope understanding, conflicts, and blockers
`phase-a/evidence/EV__phase-a2__north_star_values_and_consumer_integrity.md`	CREATE	Per-AC evidence, dispositions, semantic fixtures, baseline/final hashes, count/link/adapter/scope checks
`phase-a/RF__phase-a2__north_star_values_and_consumer_integrity.md`	CREATE	Result, deviations, verification, corrective provenance, and review handoff

Budget: 15 execution-stage files: 12 modifications and 3 new files. Estimated ≤1,000 changed LOC. Within configured limits of 30 files, 15 new files, 3,000 LOC, and 30 modified files. Root Task Board transitions are TFW process state and do not spend the product-file budget. Later review-stage files and post-APPROVE KNOWLEDGE/TECH_DEBT updates belong to their own locked stages.

5. Acceptance Criteria¶

AC-1: Exact shared-workspace baseline is captured and preserved¶

The Executor proves what existed before work and does not claim, overwrite, stage, or commit another session's state.

[ ] ONB records the start HEAD, git status --porcelain=v2, tracked diff paths and hashes, and a hash manifest for every pre-existing untracked file.
[ ] ONB separately records the tracked blob and working SHA-256/word count of .tfw/README.md, including the pre-existing brand-image insertion.
[ ] ONB records the current TFW-60 master blob and all existing phase-a/ file hashes before modifying the current master.
[ ] Every pre-existing dirty path is classified as in-scope overlap or foreign read-only state; no unclassified path remains.
[ ] Final EV compares the start/final manifests and proves that foreign paths and foreign hunks are unchanged, except changes independently made by their owner after baseline and explicitly reconciled in the ledger.
[ ] Staging is by explicit Phase A.2 path; no broad staging command absorbs foreign changes.

Gate: Baseline and final manifests, per-path/hunk ownership ledger, git diff --binary comparison for foreign tracked state, untracked hash comparison, and staged-path census.

Evidence: N/A — repository state and hashes are directly inspectable; EV preserves the reproducible checks.

AC-2: Every approved value and Success Criterion has an explicit disposition¶

No TFW-25 or TFW-32 item may disappear by omission or be declared equivalent without a semantic mapping.

[ ] EV contains eight TFW-25 value rows and four TFW-32 Success-Criteria rows.
[ ] Each row uses exactly one of EXPLICIT RESTORE, SEMANTIC MERGE, or INTENTIONAL RETIRE and names the final canonical location.
[ ] Every merge explains the shared invariant and why the new wording is not weaker; every retirement names a concrete contradiction or obsolete assumption.
[ ] Final dispositions match the owner-approved Phase HL contract; a newly discovered material contradiction stops execution and returns to the owner.
[ ] Candor Over Flattery, Structural Enforcement, Naming Creates Behavior, Portability, and Success Criteria are not retired or left implicit.
[ ] The literal Traces Over Code slogan and “output requires no manual editing” outcome use the approved bounded dispositions rather than reappearing silently.

Gate: 12-row cardinality/enum/target check plus reviewer semantic comparison from source wording to final clause.

Evidence: N/A — the source and target texts are versioned project artifacts; EV is the disposition record.

AC-3: The current North Star remains a coherent essay within its frozen ceiling [depends: AC-2]¶

The correction improves semantic completeness without reverting the current document or turning it into a mechanical list.

[ ] The problem-led sequence and the current Philosophy of Trace → TFW methodology → realizations architecture remain intact.
[ ] The existing owner brand-image insertion is preserved exactly unless the owner independently changes it after the handoff baseline.
[ ] Values and Success Criteria are integrated at the smallest coherent points; the old essay/Values section is not restored wholesale.
[ ] Final descriptive word count is ≤4,200 using the same whitespace-delimited method, and before/final counts are recorded.
[ ] The final prose contains only meaning required by the approved dispositions and coherent integration; the larger ceiling is not used as a target, preferred band, or permission for filler.
[ ] If a high-quality correction cannot fit, execution stops and proposes a ceiling amendment; no value, boundary, or success outcome is shortened away to pass the metric.
[ ] No BoK content, implementation manual, artifact inventory, or competing North Star surface is added.

Gate: Narrative/heading review, before/final word and line counts, brand-block hash, old-section similarity check, and scope scan.

Evidence: N/A — this is a versioned document quality and scope criterion.

AC-4: Restored semantics remain domain-agnostic and evidence-bounded [depends: AC-2]¶

The North Star makes the required values and outcomes usable without reviving old overclaims.

[ ] Candor requires evidence-backed disagreement, surfaced risk, and no sycophantic agreement.
[ ] Structural Enforcement requires important gates to be represented by observable structure/state rather than prose exhortation alone.
[ ] Naming Creates Behavior connects precise terms to intended agent behavior without claiming names guarantee compliance.
[ ] Portability protects durable provider-independent context without claiming every realization is Markdown-only.
[ ] Selected Trace/continuity does not state that code or other outputs are disposable or identically regenerable.
[ ] Completeness, Honesty, and Single Source of Truth retain their mapped behaviors in the final essay.
[ ] A visible Success Criteria location includes bounded authorized resumption, material-decision traceability, verified knowledge compounding, and a complete usable inspectable result subject to human acceptance.
[ ] No automatic truth, lossless context, every-message capture, self-maintaining documentation, independent agent authority, or untested capability claim appears.

Gate: Required-clause and prohibited-claim scan followed by sentence-level semantic review against TFW-25, TFW-32, current NS1–NS3, and master DoF.

Evidence: N/A — semantic inspection is reproducible from cited project sources.

AC-5: All active consumers point to real current meaning [depends: AC-3, AC-4]¶

Current normative files and the current TFW-60 contract are repaired; immutable history is not normalized by search-and-replace.

[ ] .tfw/glossary.md PV priority 0 names the full North Star purpose/principles/non-goals locus and priority 1 names actual current methodology-value clauses.
[ ] .tfw/templates/HL.md Project North Star example points to current sections/anchors, not § Values and Principles.
[ ] .tfw/conventions.md describes verification as real-item plus semantic relevance, not link existence alone.
[ ] TFW-60 header and §7.2 replace deleted-section names with current anchors and preserve the intended Structural Enforcement, Single Source of Truth, Portability, Trace, resumption, and knowledge-compounding meanings.
[ ] A repository census classifies every remaining stale string as immutable history, current knowledge/debt pending post-review correction, or a defect; no unclassified active occurrence remains.
[ ] No frozen TFW-60 claim or parallel Phase A file changes.

Gate: Active-consumer/history census, anchor resolver, TFW-60 frozen-section diff check, and semantic item-to-clause matrix.

Evidence: N/A — active paths and historical paths are inspectable repository facts.

AC-6: `/tfw-plan` has a mandatory semantic Project Values path [depends: AC-5]¶

Planning cannot skip priority 0 or cite priority 0/1 by file name alone.

[ ] Canonical plan.md requires full scans of priorities 0–4 and relevance skims of priorities 5–7.
[ ] Priority 0 and priority 1 are recorded as distinct semantic items even when they share .tfw/README.md.
[ ] HL §7.2 citations name the clause/item read and explain its concrete application; a file-only citation is insufficient.
[ ] templates/HL.md carries the same scan range and semantic requirement.
[ ] Explicit N/A remains possible only after the required scan and includes a reason.
[ ] Claude and Antigravity tfw-plan copies are byte-identical to canonical plan.md.

Gate: Source-text inspection, a fixture planning scan that separately cites priority 0 and 1, and byte hashes for both installed copies.

Evidence: N/A — workflow text and fixture outputs are local deterministic checks.

AC-7: `/tfw-review` verifies Project Values relevance, not only resolution [depends: AC-5]¶

Review must detect the exact failure that Phase A REVIEW accepted.

[ ] Canonical review.md requires the Reviewer to scan priorities 0–4 in full and 5–7 by relevance during Verify.
[ ] templates/review/verify.md checks link resolution, item existence, quoted/claimed semantic match, and relevance to the application asserted in HL §7.2/ONB §7.
[ ] Priority 0 citations are checked against purpose/principle/non-goal clauses; priority 1 citations are checked against actual methodology-value clauses.
[ ] A citation that resolves to the wrong or irrelevant section is a discrepancy and triggers the existing escalation to 100% verification.
[ ] The verification checkpoint records total, resolved, semantically verified, irrelevant, and hallucinated counts.
[ ] Claude and Antigravity tfw-review copies are byte-identical to canonical review.md.

Gate: Negative fixture reproducing the old § Values and Principles citation and positive fixture using current clauses; review source/copy hashes; semantic-count check.

Evidence: N/A — the negative and positive fixtures directly exercise the Markdown workflow contract.

AC-8: Corrective provenance preserves the old APPROVE and names its acceptance defect¶

The trace explains how correct local verification produced an incomplete project result.

[ ] Master A7, Phase HL, ONB, EV, and RF state that prior Phase A REVIEW APPROVE remains unchanged.
[ ] They identify the exact missed boundary: framework consumers were out of TS scope and the Reviewer treated resolving file pointers as sufficient.
[ ] Old RF/REVIEW/stage files and TFW-25/TFW-32 artifacts have zero diff.
[ ] The new RF never describes Phase A as wholly failed; it distinguishes a good essay from incomplete consumer/acceptance coverage.
[ ] Active-consumer corrections cite the old review observation rather than erasing it.
[ ] The explicit owner A7/HL/TS ruling, including the 4,200-word ceiling, is recorded at the re-freeze baseline; no broader authority is inferred from it.

Gate: Historical path diff, required provenance statements, and owner-verdict record check.

Evidence: N/A — provenance is established by the immutable artifacts and amendment log.

AC-9: Integrity, scope, and copy checks pass on the exact final state [depends: AC-1, AC-5, AC-6, AC-7]¶

All final claims are retaken after edits rather than copied from an intermediate state.

[ ] Every local Markdown path and named anchor introduced or modified resolves on the final tree.
[ ] Canonical workflow and installed-copy byte comparisons pass for plan and review.
[ ] Final word count, active stale-reference census, foreign-state manifest, and TFW-60 frozen diff are captured by one final verification pass immediately before RF.
[ ] EV stores raw command/output summaries sufficient to reproduce counts and hashes; RF reports the same numbers.
[ ] git diff --check passes and strict UTF-8/mojibake checks pass for modified text files.
[ ] Production diff contains only the 12 planned current consumers plus normal TFW-55 process traces/status; no BoK or unrelated file enters the result.

Gate: Exact-final verification bundle and EV↔RF number/hash consistency check.

Evidence: N/A — repository integrity checks are synthetic local verification, recorded in the required EV.

AC-10: Execution, review, returns, and knowledge closure remain role-separated¶

Approval does not collapse later roles or allow an endless corrective loop.

[ ] Executor creates a new Phase A.2 ONB before production, then a new EV and RF; it does not write REVIEW or run /tfw-docs.
[ ] A separate Reviewer creates new phase-a/review/phase-a2/map.md, phase-a/review/phase-a2/verify.md, phase-a/review/phase-a2/judge.md, and REVIEW__phase-a2__north_star_values_and_consumer_integrity.md.
[ ] The Reviewer explicitly reruns AC-2, AC-5, AC-6, and AC-7 semantic checks rather than trusting RF.
[ ] At most three formal REVISE returns are allowed; after the third, the next action is owner escalation, not a fourth automatic return.
[ ] /tfw-docs runs only after REVIEW APPROVE to correct current KNOWLEDGE wording and resolve/reclassify TD-166; no BoK work starts.
[ ] Task Board/history continues to show the prior Phase A APPROVE and the corrective A.2 chain separately.

Gate: Artifact/session lineage, review-return counter, REVIEW verdict, and post-review workflow routing.

Evidence: DEFERRED — formal Reviewer and post-review Coordinator artifacts cannot exist during Executor handoff; RF must name the required next command without claiming completion.

Evidence Artifact¶

File	Description
`phase-a/evidence/EV__phase-a2__north_star_values_and_consumer_integrity.md`	Exact handoff/final manifests, 12-row disposition ledger, current/historical consumer census, semantic source-to-target maps, word/anchor/link checks, plan/review positive and negative fixtures, adapter hashes, TFW-60 frozen/parallel preservation, per-AC verdicts, and raw final verification summaries

6. Technical Guidance¶

Source order¶

Master HL at e3e7b33, proposed A7, and the explicit owner verdict when recorded.
This Phase HL and TS.
Original Phase A TS, RF, REVIEW, review/judge.md, and current .tfw/README.md; use them to preserve the good result and identify the missed boundary.
TFW-25 TS/RF for the eight approved values and TFW-32 Phase D improvement spec/current KNOWLEDGE legacy row for Success Criteria.
Current glossary, conventions, HL template, plan/review workflows, verify template, full-copy mapping, and TFW-60 master HL.
The complete handoff dirty-state manifest; latest file content always wins over an older planning snapshot unless it conflicts with the approved contract.

Disposition semantics¶

EXPLICIT RESTORE means the value/outcome is visibly named in the current canonical essay.
SEMANTIC MERGE means the invariant survives in an identified current clause; the ledger must prove equivalence and losslessness.
INTENTIONAL RETIRE means the old wording conflicts with the current North Star or evidence boundary; the ledger names the conflict and any bounded replacement.
Do not infer a fourth state such as “already implied.” Implication without a target and reason is the silent-loss defect.

Dirty-state integration¶

Capture before editing; compare again immediately before RF.
Edit source workflows first, then synchronize their installed copies with the repository's full-copy rule.
Apply hunk-level changes to the latest TFW-60 master. Never restore that file from the planning snapshot.
Stage only explicit Phase A.2 paths. If an in-scope file has a foreign overlapping hunk whose owner cannot be distinguished, stop and escalate rather than choosing ownership.
The existing .tfw/README.md image insertion is input, not Phase A.2 authorship.

North Star editing boundary¶

Preserve the current narrative, definition, human/agent boundary, NS1–NS3, proportional realization section, and authority routing.
The 2,645-word measured planning headroom is safety margin, not an allocation target. Prefer semantic merge and precise naming; add only necessary meaning and never expand merely because capacity exists.
If the bounded dispositions cannot fit coherently within 4,200 words, stop and return for a new owner-approved amendment; delegated authority cannot accept the overrun.

Active consumer versus history¶

Current framework instructions/templates/glossary and the current active TFW-60 master are consumers.
Completed task traces and changelog entries are historical evidence. They may contain retired strings and must not be normalized.
KNOWLEDGE D44/current task index and TD-166 are current indexes, but their correction belongs after REVIEW under /tfw-docs/debt closure, not the Executor diff.

7. Definition of Failure¶

❌ Any of the eight TFW-25 values or four TFW-32 Success-Criteria outcomes lacks an explicit disposition, target, and reason.
❌ Candor, Structural Enforcement, Naming, Portability, or Success Criteria remains absent/implicit without a new explicit owner ruling.
❌ Traces Over Code, code disposability, deterministic regeneration, no-manual-editing absolutism, lossless context, automatic truth, or independent agent authority returns unbounded.
❌ The North Star ceases to be a coherent problem-led essay, exceeds 4,200 descriptive whitespace-delimited words without a new owner-approved amendment, expands mechanically toward the ceiling, or loses the owner brand-image insertion.
❌ A current citation passes because its file exists while the claimed clause is missing, irrelevant, or semantically different.
❌ /tfw-plan still omits PV priority 0, or /tfw-review verifies only link/item existence without priority 0/1 semantic relevance.
❌ Any active current consumer named in §4 retains an unclassified stale reference after execution.
❌ An immutable historical task trace, old verdict, TFW-60 frozen claim/parallel Phase A file, root public guide, BoK, or unrelated path changes.
❌ Any pre-existing foreign dirty hunk/file is lost, claimed, staged, or committed by Phase A.2.
❌ Canonical workflows and installed full copies differ, final counts come from an intermediate state, or EV/RF hashes/numbers disagree.
❌ Executor writes REVIEW or runs /tfw-docs; Reviewer and Executor are the same session; a fourth formal return begins without owner decision.
❌ The explicit A7 approval is stretched beyond the approved dispositions, active-consumer scope, 4,200-word maximum, role sequence, or dirty-state preservation contract.

8. Phase Risks¶

Risk	Mitigation
Restoring old labels recreates old doctrine	Source-to-target disposition ledger and prohibited-claim scan
Essay quality is sacrificed to completeness or count	Semantic merges, coherent-essay review, 4,200-word maximum, no-filler check, and stop-for-amendment gate
PV priority 0/1 same-file structure is collapsed	Separate semantic citation rows and positive/negative fixtures
Historical strings create false “remaining defect” counts	Active/history classification before any sweep and final unclassified count = 0
Parallel TFW-60 work changes during execution	Re-capture at handoff, latest-file hunk integration, frozen-section diff, no phase-a edits
Full-copy adapters drift	Canonical-source-first edit and exact byte comparison for four installed copies
Corrective review repeats anchor-only acceptance	AC-7 negative fixture plus independent re-run in formal review
Review churn becomes unbounded	Three-return ceiling and mandatory owner escalation

9. Cross-Phase Modifications¶

File	Also involved in	Coordination note
`.tfw/README.md`	Original Phase A approved result; owner dirty brand insertion	Preserve current essay and foreign insertion; Phase A.2 owns only approved semantic integration
`tasks/TFW-60__conflict_resistant_shared_workspace/HL-TFW-60__conflict_resistant_shared_workspace.md`	Active TFW-60 planning and frozen contract	Header and §7.2 only; latest file/hunks at execution; no frozen claim or phase-a draft edit
`.tfw/workflows/plan.md`, `.tfw/workflows/review.md`	Installed Claude/Antigravity full copies	Canonical source owns meaning; installed copies must be byte-identical
`README.md` Task Board	Parallel TFW-60 status and normal TFW-55 transitions	Preserve foreign TFW-60 row exactly; only normal TFW-55 A.2 status/link hunk belongs to this task
`KNOWLEDGE.md`, `TECH_DEBT.md`	D44/current TFW-55 record and TD-166	Read during execution/review; modify only after REVIEW APPROVE through `/tfw-docs`/debt closure

TS — TFW-55 / Phase A.2: North Star Values and Consumer Integrity | 2026-08-26