Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
44 commits
Select commit Hold shift + click to select a range
3fa03d2
docs(retro): retain PR 34 delivery evidence
teslamint Sep 8, 2026
45c6886
docs(release-loop): clarify legacy V1 handoff evidence boundaries
teslamint Sep 19, 2026
2e23ba9
chore(release-loop): retire completed fuzzing ledger
teslamint Sep 19, 2026
7ccd260
test(evals): capture compound behavior coverage
teslamint Sep 19, 2026
4e91fa6
chore(repo): ignore generated evaluation and bytecode artifacts
teslamint Sep 19, 2026
37cc5a3
chore(omp): register compound-loop skill manifest
teslamint Sep 19, 2026
d28a9a5
docs(design): capture auto planning decision contract
teslamint Sep 19, 2026
50b0423
docs(design): approve auto planning contract
teslamint Sep 19, 2026
4fb88e2
docs(plan): record inventory recheck contradiction
teslamint Sep 19, 2026
15a97c7
docs(plan): create auto decision contract plan
teslamint Sep 19, 2026
60238af
docs(plan): approve auto decision contract plan
teslamint Sep 19, 2026
4ab1a46
docs(plan): record auto contract correction
teslamint Sep 19, 2026
55a4b2b
docs(plan): supersede incomplete auto plan
teslamint Sep 19, 2026
237522c
docs(plan): approve corrected auto contract plan
teslamint Sep 19, 2026
c061f23
docs(design): correct auto trade-off evidence
teslamint Sep 19, 2026
b04b5d3
docs(plan): supersede plan with auto-mode origin
teslamint Sep 19, 2026
c0323c2
docs(design): approve corrected auto contract
teslamint Sep 19, 2026
6d1c2ba
docs(plan): align final plan with corrected design
teslamint Sep 19, 2026
680c6c5
docs(plan): approve final auto contract plan
teslamint Sep 19, 2026
b383712
feat(auto): require assumptions and trade-offs
teslamint Sep 19, 2026
0c0a536
docs(plan): require explicit closure sections
teslamint Sep 19, 2026
7a5c447
docs(plan): approve closure contract
teslamint Sep 19, 2026
dbc4f6f
fix(auto): close artifact contract review gaps
teslamint Sep 19, 2026
211142b
fix(auto): require executable success evidence
teslamint Sep 19, 2026
d7d3f56
fix(auto): validate evidence fields and plan rechecks
teslamint Sep 19, 2026
4fd7a35
docs(release-loop): bind auto coverage to ledger outputs
teslamint Sep 20, 2026
09f10ac
docs(plan): approve auto coverage successor
teslamint Sep 20, 2026
bdb13c3
docs(spec): assert auto ledger mode in coverage
teslamint Sep 20, 2026
8b1c3ad
docs(release-loop): preserve final coverage authority
teslamint Sep 20, 2026
cdf01ef
fix(auto): preserve plan approval history and ledger controls
teslamint Sep 20, 2026
5551dff
docs(plan): approve final authority successor
teslamint Sep 20, 2026
a4e5f05
docs(retro): close auto coverage release loop
teslamint Sep 20, 2026
b4ff2c1
Release v0.11.0-dev.4
teslamint Sep 20, 2026
095c955
fix(auto): close auto-artifact validator bypasses
teslamint Nov 5, 2026
7fedc71
fix(skills): run auto validators from the loaded skill root
teslamint Nov 5, 2026
a49977d
fix(evals): let scenarios do what their graders require
teslamint Nov 5, 2026
6dd6958
docs(solutions): stop recovery where no packaged re-review exists
teslamint Nov 5, 2026
1810a9e
chore(release): align package.json with 0.11.0-dev.4
teslamint Nov 5, 2026
2b79128
fix(auto): reject only negations of the negative-control phrase
teslamint Nov 5, 2026
bb5809c
docs(skills): state the headings the auto validators require
teslamint Nov 5, 2026
b82bf07
fix(auto): treat optional as negation only when it modifies the phrase
teslamint Nov 5, 2026
6bf592a
fix(auto): restore modal negations and keep negated qualifiers valid
teslamint Nov 5, 2026
05ae14e
fix(auto)!: require a Negative controls: marker in Success Criteria
teslamint Nov 5, 2026
c243893
fix(auto): reject Markdown-wrapped negative-control placeholders
teslamint Nov 5, 2026
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion .claude-plugin/plugin.json
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
{
"name": "compound-loop",
"version": "0.11.0-dev.3",
"version": "0.11.0-dev.4",
"description": "Cross-harness engineering lifecycle skills: design, plan, implement, review, ship, retrospect, and compound knowledge. Distilled from superpowers, compound-engineering, and release-loop.",
"author": {
"name": "Jaehoon You"
Expand Down
2 changes: 1 addition & 1 deletion .codex-plugin/plugin.json
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
{
"name": "compound-loop",
"version": "0.11.0-dev.3",
"version": "0.11.0-dev.4",
"description": "Cross-harness engineering lifecycle skills: design, plan, implement, review, ship, retrospect, and compound knowledge. Distilled from superpowers, compound-engineering, and release-loop.",
"license": "MIT",
"skills": "./skills/",
Expand Down
3 changes: 3 additions & 0 deletions .gitignore
Original file line number Diff line number Diff line change
Expand Up @@ -17,4 +17,7 @@
# local state (machine-local, never commit)
.entirecontext/
.omc/

# generated artifacts
/evals/results/
__pycache__/
143 changes: 0 additions & 143 deletions .release-loop/progress.md

This file was deleted.

14 changes: 14 additions & 0 deletions CHANGELOG.md
Original file line number Diff line number Diff line change
Expand Up @@ -2,6 +2,20 @@

All notable changes to this project are documented in this file.

## [0.11.0-dev.4] - 2026-09-20

### Added
- Governed recovery for eligible legacy `archived-incomplete` runs, with pinned backup, authority audit, verified restore, and completion artifacts.
- Legacy `.release-loop/v1` evidence now transfers, resumes after partial directory copies, rejects unsafe ownership or destination state, and archives with the run.
- `release-loop --auto` specs and plans now record explicit assumptions, trade-offs, measured costs, and closure evidence, with standalone validators and negative-control coverage.

### Changed
- Coverage checks bind their production population and mutation denominator to the authoritative release-loop ledger, then report validator, downstream, and stacked recall separately.
- Legacy handoff and archive contracts now preserve symlink, collision, malformed-ownership, and fixture-identity rejection evidence.

### Fixed
- CI now rejects prohibited fixture author or committer identities before merge.

## [0.11.0] - 2026-08-22

### Added
Expand Down
3 changes: 2 additions & 1 deletion ROADMAP.md
Original file line number Diff line number Diff line change
Expand Up @@ -66,7 +66,8 @@ Open items registered by retro docs (`docs/retros/`), pushed here per `retrospec
| Post-merge fix claims retain summarized results but not publisher-owned raw failure and RED/GREEN traces | 2026-08-25 run-artifact-integrity retro (T3) | P2 | Next merged-result failure that creates a follow-up — publish bounded raw output and exact environment observations before citing root cause |
| Two spec-declared test case names (`legacy_handoff_source_changed`, `legacy_handoff_complete_rerun`) were never implemented; add them to close the SC3/SC4 measurement gap | 2026-08-25 legacy-handoff-contract retro (T2) | P2 | Next change to legacy handoff test cases or spec measurement contract |
| Add a mid-directory-copy fault injection (`handoff-after-copy-one-file`) to exercise partial-directory recovery directly | 2026-08-25 legacy-handoff-contract retro (T3) | P3 | Next change to handoff fault injection or recovery code |
| Structured review events must be populated when `completeness: exact` — either use the event system or downgrade to `partial` | 2026-08-25 legacy-handoff-contract retro | P3 | Next review phase that sets `completeness: exact` |
| ~~Structured review events must be populated when `completeness: exact` — either use the event system or downgrade to `partial`~~ **Done** | 2026-08-25 legacy-handoff-contract retro; resolved 2026-09-20 auto-plan-assumption-tradeoffs retro (T1) | P3 | Resolved: events 1–3 use publisher-owned wrappers, and the ledger explicitly retains `completeness: partial` rather than claiming exact historical coverage |
| Review re-review verification must be executable rather than manual ledger editing | 2026-09-19 auto-plan-assumption-tradeoffs retro (T2) | P2 | Next change to reviewing or implementing review-event operations — expose a packaged `verify_re_review` path that validates source/result inventories and writes fixed dispositions atomically |

## Non-goals (re-affirmed)

Expand Down
52 changes: 52 additions & 0 deletions docs/deviations/2026-09-19-auto-design-contract-correction-003.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,52 @@
# Deviation Addendum 003: Auto-design Trade-off Cost Correction

_Recorded 2026-09-19 before implementation resumed._

## Original contract

The approved design at
`docs/specs/2026-09-19-auto-plan-assumption-tradeoffs-design.md` defined the
selected auto-only structured contract and its four-column trade-off table.

## Discovered contradiction

The first standalone validator run against the approved design found that the
selected row's `Measured cost` cell had no numeric quantity. The artifact
therefore failed the active auto-mode contract even though its rejected rows had
numeric values.

## Necessity

The selected option must name a quantified, attributed cost. The approved
artifact remains unchanged as historical evidence. A successor draft records
the measured count of two standalone validator copies and the command that
observed it at `2026-09-19T14:05:58Z`.

## Observable behavior

No runtime behavior changes. The successor design only repairs the evidence
shape required before Architecture and Success Criteria. It does not widen the
selected scope, interactive gates, migration boundary, or publication boundary.

## Safety and consent boundaries

The correction is local and reversible. It changes no protected branch, ledger
authority, credential, network target, or external service. The successor must
receive a fresh human Design approval before planning continues.

## Verification changes

- The successor design passes `validate-auto-artifact.py --mode auto`.
- Its selected and rejected trade-off rows contain numeric measured costs.
- The focused harness continues to require numeric costs in every trade-off row.
- The original approved design remains a historical artifact and is not used as
the implementation origin after the successor approval.

## Traceability

- Original approved design: commit `44ae6f2c938624bd29ba1c92ef5b2322bd5b1df6`.
- Finding: first standalone validator run on 2026-09-19.
- Successor draft: `docs/specs/2026-09-19-auto-plan-assumption-tradeoffs-correction-design.md`.
- Evidence command: validator-copy count at `2026-09-19T14:05:58Z`.
- Authority for this artifact shape:
`docs/solutions/workflow-issues/review-introduced-state-machine-deviation.md`.
65 changes: 65 additions & 0 deletions docs/deviations/2026-09-19-auto-plan-assumption-inventory-001.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,65 @@
# Deviation Addendum 001: Auto-plan Inventory Recheck

_Recorded 2026-09-19 during Plan preparation._

## Original contract

The approved design at
`docs/specs/2026-09-19-auto-plan-assumption-tradeoffs-design.md`
records a live assumption that the repository contains `27` tracked spec files
and `29` tracked plan files. The evidence command counts both glob sets with
`git ls-files`.

## Discovered contradiction

The planning Assumption Recheck reran that command exactly at
`2026-09-19T13:50:38Z`. It returned `spec_count=28` and `plan_count=29`.
The additional spec is the approved origin artifact itself, which this release
loop committed after the design measurement. The approved count was a
pre-feature inventory, while the identical command now includes the new spec.

A bounded follow-up command at `2026-09-19T13:51:10Z` returned
`tracked_spec_count=28; excluding_origin=27`, using the exact origin path as the
only exclusion. The pre-feature claim therefore remains reproducible, but the
unqualified recheck command is no longer a stable comparison after artifact
creation.

## Necessity

Planning cannot mark the retained claim as an unqualified match after the
identical command changes. The approved spec must remain unchanged, and the
plan must preserve the contradiction rather than silently reinterpret the
measurement. The addendum records the artifact-count lifecycle and supplies a
bounded comparison for this plan.

## Observable behavior

No product, skill, validator, terminal-output, or compatibility behavior
changes. The planning record gains one explicit contradiction classification and
one origin-aware evidence command. Historical specs and plans remain untouched.

## Safety and consent boundaries

The commands read tracked paths only. They perform no network access, file
mutation, credential access, or outward action. The design approval and the
implementation-deferred stop condition remain unchanged.

## Verification changes

- The plan's Assumption Recheck records the original command as `contradiction`
and links this addendum.
- The plan records the origin-aware count as planning evidence, not as a rewrite
of the approved design measurement.
- The focused contract harness must continue to reject missing, blank, and
unquantified auto-artifact sections; this addendum does not weaken those checks.
- Repository validation remains required before the draft plan commit.

## Traceability

- Approved spec: commit `44ae6f2c938624bd29ba1c92ef5b2322bd5b1df6`.
- Contradictory retained command: Assumptions and Preconditions row 2 in the
approved spec.
- Planning recheck: first-hand run at `2026-09-19T13:50:38Z`.
- Corrected bounded evidence: origin-aware count at `2026-09-19T13:51:10Z`.
- Authority for this artifact shape:
`docs/solutions/workflow-issues/review-introduced-state-machine-deviation.md`.
58 changes: 58 additions & 0 deletions docs/deviations/2026-09-19-auto-plan-contract-correction-002.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,58 @@
# Deviation Addendum 002: Auto-plan Assumption Contract Correction

_Recorded 2026-09-19 before implementation resumed._

## Original contract

The approved plan at
`docs/plans/2026-09-19-001-feat-auto-plan-assumption-tradeoffs-plan.md`
was intended to carry the approved auto artifact contract into four
implementation units. Its `Trade-offs` section has the required four-column
table, and its `Assumption Recheck` section records planning outcomes.

## Discovered contradiction

A pre-implementation contract review found that the approved plan's
`Assumptions and Preconditions` section used the three-column planning
`Assumption Recheck` shape instead of the required five-field evidence table:
claim, exact rerunnable command, observation timestamp, concise result, and
evidence source. The active auto-mode requirement applies to plans as well as
specs, so the plan did not satisfy its own front-loaded assumption contract.

## Necessity

Implementation cannot start from an approved plan that fails the active
artifact contract. The plan body must gain the missing five-field table while
retaining the separate three-column `Assumption Recheck` authority required by
`plan/v1`. The plan is corrected and re-sealed before any implementation file
changes.

## Observable behavior

No runtime or interactive behavior changes. The correction changes only the
planning artifact's evidence structure and adds no implementation unit,
validator rule, dependency, or outward action.

## Safety and consent boundaries

The correction is local and reversible. It changes no skill, script, release
ledger authority, credential, network target, or protected branch. Auto plan
approval remains bounded by the validator, body seal, and exact progress ledger.

## Verification changes

- The corrected plan contains a five-field `Assumptions and Preconditions`
table before Architecture notes.
- The corrected plan retains a separate `Assumption Recheck` section with one
outcome per retained claim.
- `validate-plan-frontmatter.py` recomputes the corrected body seal.
- The implementation harness must validate the same five-field requirement for
auto specs and plans and must fail when any field is blank or missing.

## Traceability

- Original approved plan: commit `d17ee7e59f9a7e95fc8e24de62c47b420d71195c`.
- Finding: pre-implementation contract review on 2026-09-19.
- Affected artifact: `docs/plans/2026-09-19-001-feat-auto-plan-assumption-tradeoffs-plan.md`.
- Authority for this artifact shape:
`docs/solutions/workflow-issues/review-introduced-state-machine-deviation.md`.
57 changes: 57 additions & 0 deletions docs/deviations/2026-09-19-auto-plan-sections-correction-004.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,57 @@
---
schema: deviation/v1
title: Auto plan closure sections correction
date: 2026-09-19
status: accepted
---

# Auto plan closure sections correction

## Original contract

The active auto-authoring rule requires every spec and plan to state assumptions,
trade-offs, non-goals, residual risks, and measurable success criteria. Each
numeric claim must carry rerunnable command evidence.

## Discovered contradiction

Approved plan `2026-09-19-003-feat-auto-plan-assumption-tradeoffs-final-plan.md`
contained the required assumptions and trade-offs tables, but it did not contain
explicit `Non-goals`, `Residual Risks`, or `Success Criteria` sections. Its
`Deferred to Follow-Up Work`, `Open unknowns`, and scenario tables did not satisfy
the active section contract.

## Necessity

The plan is an implementation authority. Leaving the omission would allow the
implementation to pass a weaker contract than the one the user requested.
The approved body remains immutable, so a successor plan is required.

## Observable behavior

Plan `2026-09-19-003-feat-auto-plan-assumption-tradeoffs-final-plan.md` is marked
`superseded`. Successor plan
`2026-09-19-004-feat-auto-plan-assumption-tradeoffs-contract-plan.md` carries
`mode: auto`, the same approved origin, explicit closure sections, and the same
implementation units.

## Safety and consent

This correction changes no runtime behavior and modifies no historical artifact.
It only makes the implementation authority satisfy the active authoring contract.
The successor plan receives the automatic plan gate because the ledger invocation
is `--auto`; no outward publish, push, or merge is authorized by this addendum.

## Verification changes

The validator now rejects an auto artifact missing any closure section. The
focused harness adds negative controls for missing closure sections and empty
selected or rejected status names. It also rejects duplicate mode entries and
unsupported validator modes.

## Traceability

- Superseded plan: `docs/plans/2026-09-19-003-feat-auto-plan-assumption-tradeoffs-final-plan.md`
- Successor plan: `docs/plans/2026-09-19-004-feat-auto-plan-assumption-tradeoffs-contract-plan.md`
- Implementation contract: `skills/designing/scripts/validate-auto-artifact.py`
- Regression harness: `scripts/test-auto-artifact-contract.sh`
Original file line number Diff line number Diff line change
@@ -0,0 +1,49 @@
---
schema: deviation/v1
title: Auto coverage authority successor correction
date: 2026-09-20
status: accepted
---

# Auto coverage authority successor correction

## Original contract

The approved coverage design and sealed successor plan must remain immutable
once the plan approval seal exists. The release ledger must point to the
approved pair used by Ship.

## Discovered contradiction

A review found that the approved coverage design received a post-plan-seal
change to assert the ledger's `--auto` flag. That change made the old plan and
its design origin non-authoritative for exact-head review.

## Necessity

The change must not mutate an approved design after its plan is sealed. The
immutable authority successor preserves the final command, fixed mutation
inventory, and ledger population assertion while replacing both predecessor
artifacts through explicit successor pointers.

## Measurement

The successor design's exact Population and Coverage command ran at
`2026-09-20T02:51:37Z`. It read the two paths from the ledger, asserted
`flags == ['[--auto]']`, validated both artifacts, and compared the fixed 17
mutation labels. It returned 17/17 validator detection, 17/17 harness
observation, validator recall `1.000`, harness incremental recall `0.000`, and
stacked recall `1.000`.

## Safety and consent

This correction changes no runtime behavior. It adds no migration requirement.
The successor remains at the Design gate until fresh user approval, followed by
a new sealed plan approval and exact-head review.

## Traceability

- Superseded design: `docs/specs/2026-09-20-auto-plan-assumption-tradeoffs-coverage-design.md`
- Successor design: `docs/specs/2026-09-20-auto-plan-assumption-tradeoffs-coverage-authority-design.md`
- Superseded plan: `docs/plans/2026-09-20-005-feat-auto-plan-assumption-tradeoffs-coverage-plan.md`
- Successor plan: `docs/plans/2026-09-20-006-feat-auto-plan-assumption-tradeoffs-coverage-authority-plan.md`
50 changes: 50 additions & 0 deletions docs/deviations/2026-09-20-auto-goal-coverage-correction-005.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,50 @@
---
schema: deviation/v1
title: Auto artifact goal coverage correction
date: 2026-09-20
status: accepted
---

# Auto artifact goal coverage correction

## Original contract

The approved successor design requires executable assumptions, trade-offs,
closure sections, and negative controls for artifacts written by `--auto`.

## Discovered contradiction

The success criteria proved individual fields and mutations, but they did not
state the population covered by the goal or require a recall floor. A green
field-level harness could therefore leave broad artifact coverage unmeasured.

## Necessity

The goal covers every spec and plan produced by the two auto-authoring skills.
A design that lacks a population and measured coverage cannot prove that its
implementation attains that goal. The approved design remains immutable, so a
successor design records the coverage contract and the measured sample.

## Measurement

The rerunnable Population and Coverage command in the successor design ran at
`2026-09-20T02:38:06Z`. It validated the two artifacts named by the
authoritative ledger. It tested 17 fixed malformed mutations: the standalone
validators detected 17/17, the downstream harness observed 17/17, validator
recall was `1.000`, harness incremental recall was `0.000`, and stacked recall
was `1.000`. The harness is downstream, so zero incremental recall is expected
and is not presented as complementary coverage.

## Safety and consent

This correction changes no runtime behavior. It adds no migration requirement
and does not weaken the auto-only scope. The successor design was approved at
the Design gate after the measured coverage was reviewed.

## Traceability

- Superseded design: `docs/specs/2026-09-19-auto-plan-assumption-tradeoffs-correction-design.md`
- Successor design: `docs/specs/2026-09-20-auto-plan-assumption-tradeoffs-coverage-design.md`
- Measurement command: `bash scripts/test-auto-artifact-contract.sh`
- Implementation contract: `skills/designing/scripts/validate-auto-artifact.py`
- Regression harness: `scripts/test-auto-artifact-contract.sh`
Loading
Loading