morluto/rea

Fork 0

mirror of https://github.com/morluto/rea.git synced 2026-11-05 06:45:22 +00:00

[PR #5722] [CLOSED] feat(environment): build an environment from several recording sessions #7037

New issue

Closed

opened 2026-11-01 21:44:53 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-11-01 21:44:53 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/morluto/rea/pull/5722
Author: @earakely-scale
Created: 10/20/2026
Status: ❌ Closed

Base: main ← Head: feat/environment-sessions

📝 Commits (10+)

ad439bd feat(environment): extract the initial state a capture's reads observed
7a668b7 Merge main into feat/environment-state-view
f2c9a2e fix(environment): bound state at redacted writes; infer identity only from items
2999128 fix(environment): keep first sightings, one list pointer and linear field unions
a5c7b92 fix(environment): keep post-write values and partial lists out of the initial state
4e93547 docs(environment): scope item-shaped and deleted written paths out of documents
aba9cb1 fix(environment): count writes in flight when a value changes between reads
856c8c0 Merge main into feat/environment-state-view
55480bd chore: merge main (5fa4defc) into environment state view
9fee36d fix(environment): catch cross-tool changes across writes; read only item totals

📊 Changes

92 files changed (+16182 additions, -726 deletions)

View changed files

📝 .agents/skills/reverse-engineer-anything/references/evidence-workflows.md (+11 -1)
📝 docs/agent-environments.md (+387 -52)
📝 docs/cli.md (+35 -19)
📝 docs/mcp-contracts.md (+1 -1)
➕ docs/specs/agent-environments/recording-sessions/plan.md (+19 -0)
➕ docs/specs/agent-environments/recording-sessions/spec.md (+70 -0)
➕ docs/specs/agent-environments/recording-sessions/tasks.md (+13 -0)
➕ docs/specs/agent-environments/recording-sessions/verification.md (+95 -0)
➕ docs/specs/agent-environments/state-coverage/plan.md (+32 -0)
➕ docs/specs/agent-environments/state-coverage/spec.md (+94 -0)
➕ docs/specs/agent-environments/state-coverage/tasks.md (+16 -0)
➕ docs/specs/agent-environments/state-coverage/verification.md (+112 -0)
➕ docs/specs/agent-environments/state-serve/plan.md (+44 -0)
➕ docs/specs/agent-environments/state-serve/spec.md (+189 -0)
➕ docs/specs/agent-environments/state-serve/tasks.md (+32 -0)
➕ docs/specs/agent-environments/state-serve/verification.md (+72 -0)
➕ docs/specs/agent-environments/state-writes/plan.md (+45 -0)
➕ docs/specs/agent-environments/state-writes/spec.md (+348 -0)
➕ docs/specs/agent-environments/state-writes/tasks.md (+22 -0)
➕ docs/specs/agent-environments/state-writes/verification.md (+225 -0)

...and 72 more files

📄 Description

Summary

An environment is built from what its capture shows. When one recording writes once and never reads the result back, the write stays unserved. A call the recording never made is answered from no evidence. Recording more sessions of the same application is the natural fix, but REA reads a capture as one session: appended to the first, a second recording's reads come after the first one's writes, and its writes are replayed on top of the first one's.

This PR lets a capture hold several recording sessions, each recorded against the application restored to the same state (for example one database snapshot). Each session's reads before its own first related write show the initial state, and each session's writes are replayed from that state. A write any session confirms is served unless another session contradicts it. A plan is also now built from the first call that can build it, so a later session can supply the read the capture's own call lacked.

Part of #4933. Stacked on #5600, merged in at f2b95351; until #5600 lands, the diff against main includes it. The new work is:

session_starts on the manifest and state view requests (environmentManifestView.ts, environmentStateView.ts) and in the analysis-view tool contract;
the per-session timeline, cutoffs and change classification (EnvironmentStateAccumulator.ts, EnvironmentStateExtraction.ts, environmentStateRules.ts, environmentStateWriteCutoffs.ts);
the per-session replay and the later-call build (EnvironmentStateWriteSimulation.ts, new environmentStateWriteSession.ts);
serving with the manifest's sessions (environmentState.ts);
tests, docs and docs/specs/agent-environments/recording-sessions/.

Problem and expected behavior

{
  "source": { "kind": "inline", "evidence": { "evidence_id": "ev_…" } },
  "view": {
    "kind": "environment-manifest",
    "template_rules": ["after-uid-literal", "opaque-token", "sibling-variance"],
    "session_starts": [42]
  }
}

session_starts lists the ordinals where a later session begins, strictly ascending; for a HAR, an entry's ordinal is its index in log.entries, so a second recording appended to a first starts at the first's entry count.

Initial state:
- records are ordered by session, then by time, so sessions that overlap in time stay apart;
- each session's successful reads before its own first related write show the initial state;
- across sessions, only the later session's own writes lie between two observations, so a value two sessions' initial reads disagree on is volatile, or fields-differ if the later session wrote first;
- a limitation counts the sessions and states that their shared starting state is the caller's claim.
Write gate: each session replays on its own copy of the initial state, with its own taint and checks. Plans, confirmations and contradictions are shared. A plan whose first call had an unusable body or showed no entity is built from the next call that can build it, instead of being fixed by that first call.
Serving and export: the manifest records session_starts, so environment-serve and environment-export derive the state with the same sessions and refuse a supplied state extracted with others.
Without session_starts: nothing changes, and the echoed request stays the same, so existing manifests and states keep their Evidence ids.

Change and scope

Self-contained: request schema, the state's timeline and write gate, and the serve-time check. The manifest groups records the same way with or without sessions.
Not included:
- recording or concatenating sessions: the docs show jq for joining HARs;
- checking that sessions really started from one state, which REA cannot observe.
Scope: analysis of existing Evidence, inside the existing boundary; no target execution.

Contract and boundary impact

Semantic owner and earliest changed stage: the environment-manifest and environment-state analysis views (src/domain/analysisView/), then serving (src/application/environment/).
Contribution scope fit and linked maintainer decision: extends the views' request and the write gate; not applicable.
CLI and MCP/tool-catalog contract: inspect_analysis_view accepts an optional session_starts on the two environment views; the tool description says so. The catalog check passes.
Provider, bridge, target-format, or platform compatibility: none.
Evidence, artifact, provenance, or reconstruction contract: a view requested without sessions is byte-identical; with them, the request is echoed and a limitation is added. A plan can now be built from a later call than its first; exemplar_ordinal names that call.
Target selection, authority, upstream behavior, and cleanup details (if applicable): not applicable.
Generated metadata (docs/public/product-catalog.json), package, or installation impact: none.

Expected effects (check known effects; use documentation-only on its own).
Unchecked items do not establish absence; leave uncertainties visible.

Read existing files or captured data.
Launch or attach to a process.
Inject instrumentation or run scripts/code.
Change process memory, registers, files, or configuration.
Contact network services.
Recover protected code or work around protections.
Documentation or usability only; no target interaction.
Not sure yet; investigation or maintainer guidance needed.

Evidence and regression coverage

Tests added.

Domain (environmentStateSessions, 7), through the state view:
- a create repeated after a reset is confirmed, while the same recordings read as one session contradict it;
- a plan the first session's call cannot build is built from a later session's call;
- a write one session confirms and another contradicts is withheld;
- a later session's reads before its own writes back an item view;
- a value the sessions' initial reads disagree on is volatile, and fields-differ when the later session wrote first;
- sessions overlapping in time stay together;
- request validation.
Service (environmentState, 1): the state derived from a manifest with sessions is the state view requested with them, and a state extracted without them is refused for that manifest.

Executed evidence: six applications, three builds, held-out sessions. The six applications of the fidelity comparison (#5699) were restored from one snapshot before every recording. A driver outside REA recorded:

training: a re-run of the capture, plus one create/read/list/update/delete session per write target, served or not;
held-out: values the training never sent (two creates, updates and deletes of created and seeded entities), and a read of every seeded entity, recorded after the training.

Each build was compared with the held-out sessions by environment-compare, on a local build with #5699 and #5664 merged:

one: the re-run alone;
sessions: the re-run plus the training sessions, with session_starts;
joined: the same recordings read as one session.

Application	Steps	one	sessions	joined
NetBox	273	98.9%	98.9%	98.9%
Komga	365	94.2%	92.9%	92.9%
Mealie	310	94.2%	94.2%	94.2%
n8n	143	88.8%	89.5%	89.5%
Shlink	137	92.0%	92.0%	92.0%
Sentry	226	40.7%	51.3%	40.7%

The extra sessions supplied the evidence; what they showed was a modelling gap each time, so agreement barely moved:

Mealie: the capture's recipe create answers only a slug and is never read back, so all recipe plans were no-observation. The later-call rule now builds them from the training session. The create is then contradicted by a list that keeps its recorded members.
Komga: the training session lists after a create. That list keeps its recorded members (2 where Komga has 3), so the gate now withholds the create (side-effect) that the capture alone served. The drop is that create no longer being served.
Lists: membership is decided per tool, and each of these tools has filtered recordings, so their lists never show created entities. Per-line membership is the next fix.
Shlink: stores sent tags sorted, so its update fails the echo rule in every build.
n8n: the delete-before-archive 400 is not modelled.
Sentry: the gain is recency, not modelling. The time-bucketed views take their latest initial values from sessions recorded closer to the held-out reads. Joined reading loses those reads.

Reading the same recordings as one session never did better than sessions.

Found along the way:

A joined HAR is bounded by the HAR decoder's fixed 192 MiB heap: n8n's needed one 10.9 MB response (over the state's 256 KiB body limit) left out of the added session.
n8n's first training session carried an unrelated write from the capture's prefix and was re-recorded.

Details are in docs/specs/agent-environments/recording-sessions/verification.md.

Remaining proof gaps:

REA cannot check that sessions shared a starting state; the measurement used snapshot restores.
The held-out sessions cover each application's written collections and seeded reads, not every tool.

For evidence-bearing changes:

Observed, derived, and inferred claims remain distinguishable.
Artifact identity, source provenance, and failed attempts remain preserved.
Unsupported, incomplete, unavailable, or uncertain outcomes remain visible.

Validation performed

Run on macOS with Node v22.23.1 at e49361ca.

npm run build:cached, npm run check:fast, npm run format:check, npx knip, npm run scan:todos, npm run docs:check, npm run verify:test-discovery, npm run mcp-catalog:check — passed.
npm run jscpd — 3.46% over 3603 files.
npx vitest run --project domain --project services — 7484 passed, 5 pre-existing skips.
npm run test:prepare, then npm run test:focused -- the environment serve and export CLI boundaries, the serve and served-writes MCP boundaries, the analysis-view CLI and MCP boundaries, the compact input schema boundary and tool schema validity — 8 files, 78 passed.

Compatibility, safety, and release

Breaking changes or migration steps: none. Without session_starts, views and Evidence ids are unchanged. With the later-call rule, a plan whose first call could not be built may now be built and served from a later call.
Real Hopper/Ghidra, browser, or OS coverage: not applicable.
Package or release metadata impact: none.
Security, privacy, process, or containment review: no new inputs beyond ordinals; the verification's recordings were made without headers or cookies.

Author readiness

Human owner:

I reviewed the final diff and relevant existing code, understand the change and its edge cases, and take responsibility for the complete PR, including AI-assisted work.
I checked the evidence above, confirmed it supports the current change, and recorded meaningful validation gaps.
I declare this PR ready for initial maintainer review.

Before merge

This PR addresses a concrete problem or an agreed enhancement.
Material scope and authority questions are resolved, and any required maintainer scope decision covers the final diff.
The PR has one focused outcome and the title follows type(scope): outcome.
Related issue is linked, or the reason for not linking one is stated above.
Tests cover changed observable behavior and meaningful failure paths.
Owning docs, contracts, and generated metadata are updated where needed.
User-visible CLI/MCP changes include representative output.
I checked the final diff for secrets, unrelated cleanup, and unsupported claims.

🤖 Generated with Claude Code

_{🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.}

## 📋 Pull Request Information **Original PR:** https://github.com/morluto/rea/pull/5722 **Author:** [@earakely-scale](https://github.com/earakely-scale) **Created:** 10/20/2026 **Status:** ❌ Closed **Base:** `main` ← **Head:** `feat/environment-sessions` --- ### 📝 Commits (10+) - [`ad439bd`](https://github.com/morluto/rea/commit/ad439bd2f26c61b442dcf758c2bb149e49657929) feat(environment): extract the initial state a capture's reads observed - [`7a668b7`](https://github.com/morluto/rea/commit/7a668b73f4783b42eb0e8ce1eeff079f86d53e22) Merge main into feat/environment-state-view - [`f2c9a2e`](https://github.com/morluto/rea/commit/f2c9a2e4462a7162224bec209f21dfa2622abe39) fix(environment): bound state at redacted writes; infer identity only from items - [`2999128`](https://github.com/morluto/rea/commit/2999128baedceabc738e9a5aa69bfcfec39a55f1) fix(environment): keep first sightings, one list pointer and linear field unions - [`a5c7b92`](https://github.com/morluto/rea/commit/a5c7b9236c48094bc4af24ae055e8d67e3968477) fix(environment): keep post-write values and partial lists out of the initial state - [`4e93547`](https://github.com/morluto/rea/commit/4e93547c846444e891e2a2267425c316e6ee5218) docs(environment): scope item-shaped and deleted written paths out of documents - [`aba9cb1`](https://github.com/morluto/rea/commit/aba9cb15c8836b7592a51743d79bceab4dfd99af) fix(environment): count writes in flight when a value changes between reads - [`856c8c0`](https://github.com/morluto/rea/commit/856c8c0b50ec11e4ba96cb6172b7cfe01ef6829f) Merge main into feat/environment-state-view - [`55480bd`](https://github.com/morluto/rea/commit/55480bdf7e3275e3be8903e53baabe28be81a9a9) chore: merge main (5fa4defc) into environment state view - [`9fee36d`](https://github.com/morluto/rea/commit/9fee36d3155145fa8e568b34713e0e68c7675216) fix(environment): catch cross-tool changes across writes; read only item totals ### 📊 Changes **92 files changed** (+16182 additions, -726 deletions) <details> <summary>View changed files</summary> 📝 `.agents/skills/reverse-engineer-anything/references/evidence-workflows.md` (+11 -1) 📝 `docs/agent-environments.md` (+387 -52) 📝 `docs/cli.md` (+35 -19) 📝 `docs/mcp-contracts.md` (+1 -1) ➕ `docs/specs/agent-environments/recording-sessions/plan.md` (+19 -0) ➕ `docs/specs/agent-environments/recording-sessions/spec.md` (+70 -0) ➕ `docs/specs/agent-environments/recording-sessions/tasks.md` (+13 -0) ➕ `docs/specs/agent-environments/recording-sessions/verification.md` (+95 -0) ➕ `docs/specs/agent-environments/state-coverage/plan.md` (+32 -0) ➕ `docs/specs/agent-environments/state-coverage/spec.md` (+94 -0) ➕ `docs/specs/agent-environments/state-coverage/tasks.md` (+16 -0) ➕ `docs/specs/agent-environments/state-coverage/verification.md` (+112 -0) ➕ `docs/specs/agent-environments/state-serve/plan.md` (+44 -0) ➕ `docs/specs/agent-environments/state-serve/spec.md` (+189 -0) ➕ `docs/specs/agent-environments/state-serve/tasks.md` (+32 -0) ➕ `docs/specs/agent-environments/state-serve/verification.md` (+72 -0) ➕ `docs/specs/agent-environments/state-writes/plan.md` (+45 -0) ➕ `docs/specs/agent-environments/state-writes/spec.md` (+348 -0) ➕ `docs/specs/agent-environments/state-writes/tasks.md` (+22 -0) ➕ `docs/specs/agent-environments/state-writes/verification.md` (+225 -0) _...and 72 more files_ </details> ### 📄 Description ## Summary An environment is built from what its capture shows. When one recording writes once and never reads the result back, the write stays unserved. A call the recording never made is answered from no evidence. Recording more sessions of the same application is the natural fix, but REA reads a capture as one session: appended to the first, a second recording's reads come after the first one's writes, and its writes are replayed on top of the first one's. This PR lets a capture hold several **recording sessions**, each recorded against the application restored to the same state (for example one database snapshot). Each session's reads before its own first related write show the initial state, and each session's writes are replayed from that state. A write any session confirms is served unless another session contradicts it. A plan is also now built from the first call that can build it, so a later session can supply the read the capture's own call lacked. Part of #4933. **Stacked on #5600**, merged in at `f2b95351`; until #5600 lands, the diff against main includes it. The new work is: - `session_starts` on the manifest and state view requests (`environmentManifestView.ts`, `environmentStateView.ts`) and in the analysis-view tool contract; - the per-session timeline, cutoffs and change classification (`EnvironmentStateAccumulator.ts`, `EnvironmentStateExtraction.ts`, `environmentStateRules.ts`, `environmentStateWriteCutoffs.ts`); - the per-session replay and the later-call build (`EnvironmentStateWriteSimulation.ts`, new `environmentStateWriteSession.ts`); - serving with the manifest's sessions (`environmentState.ts`); - tests, docs and `docs/specs/agent-environments/recording-sessions/`. ## Problem and expected behavior ```json { "source": { "kind": "inline", "evidence": { "evidence_id": "ev_…" } }, "view": { "kind": "environment-manifest", "template_rules": ["after-uid-literal", "opaque-token", "sibling-variance"], "session_starts": [42] } } ``` `session_starts` lists the ordinals where a later session begins, strictly ascending; for a HAR, an entry's ordinal is its index in `log.entries`, so a second recording appended to a first starts at the first's entry count. - **Initial state:** - records are ordered by session, then by time, so sessions that overlap in time stay apart; - each session's successful reads before its own first related write show the initial state; - across sessions, only the later session's own writes lie between two observations, so a value two sessions' initial reads disagree on is volatile, or `fields-differ` if the later session wrote first; - a limitation counts the sessions and states that their shared starting state is the caller's claim. - **Write gate:** each session replays on its own copy of the initial state, with its own taint and checks. Plans, confirmations and contradictions are shared. A plan whose first call had an unusable body or showed no entity is built from the next call that can build it, instead of being fixed by that first call. - **Serving and export:** the manifest records `session_starts`, so `environment-serve` and `environment-export` derive the state with the same sessions and refuse a supplied state extracted with others. - **Without `session_starts`:** nothing changes, and the echoed request stays the same, so existing manifests and states keep their Evidence ids. ## Change and scope - **Self-contained:** request schema, the state's timeline and write gate, and the serve-time check. The manifest groups records the same way with or without sessions. - **Not included:** - recording or concatenating sessions: the docs show `jq` for joining HARs; - checking that sessions really started from one state, which REA cannot observe. - **Scope:** analysis of existing Evidence, inside the existing boundary; no target execution. ## Contract and boundary impact - Semantic owner and earliest changed stage: the environment-manifest and environment-state analysis views (`src/domain/analysisView/`), then serving (`src/application/environment/`). - Contribution scope fit and linked maintainer decision: extends the views' request and the write gate; not applicable. - CLI and MCP/tool-catalog contract: `inspect_analysis_view` accepts an optional `session_starts` on the two environment views; the tool description says so. The catalog check passes. - Provider, bridge, target-format, or platform compatibility: none. - Evidence, artifact, provenance, or reconstruction contract: a view requested without sessions is byte-identical; with them, the request is echoed and a limitation is added. A plan can now be built from a later call than its first; `exemplar_ordinal` names that call. - Target selection, authority, upstream behavior, and cleanup details (if applicable): not applicable. - Generated metadata (`docs/public/product-catalog.json`), package, or installation impact: none. Expected effects (check known effects; use documentation-only on its own). Unchecked items do not establish absence; leave uncertainties visible. - [x] Read existing files or captured data. - [ ] Launch or attach to a process. - [ ] Inject instrumentation or run scripts/code. - [ ] Change process memory, registers, files, or configuration. - [ ] Contact network services. - [ ] Recover protected code or work around protections. - [ ] Documentation or usability only; no target interaction. - [ ] Not sure yet; investigation or maintainer guidance needed. ## Evidence and regression coverage **Tests added.** - Domain (`environmentStateSessions`, 7), through the state view: - a create repeated after a reset is confirmed, while the same recordings read as one session contradict it; - a plan the first session's call cannot build is built from a later session's call; - a write one session confirms and another contradicts is withheld; - a later session's reads before its own writes back an item view; - a value the sessions' initial reads disagree on is volatile, and `fields-differ` when the later session wrote first; - sessions overlapping in time stay together; - request validation. - Service (`environmentState`, 1): the state derived from a manifest with sessions is the state view requested with them, and a state extracted without them is refused for that manifest. **Executed evidence: six applications, three builds, held-out sessions.** The six applications of the fidelity comparison (#5699) were restored from one snapshot before every recording. A driver outside REA recorded: - training: a re-run of the capture, plus one create/read/list/update/delete session per write target, served or not; - held-out: values the training never sent (two creates, updates and deletes of created and seeded entities), and a read of every seeded entity, recorded after the training. Each build was compared with the held-out sessions by `environment-compare`, on a local build with #5699 and #5664 merged: - **one:** the re-run alone; - **sessions:** the re-run plus the training sessions, with `session_starts`; - **joined:** the same recordings read as one session. | Application | Steps | one | sessions | joined | | --- | ---: | ---: | ---: | ---: | | NetBox | 273 | 98.9% | 98.9% | 98.9% | | Komga | 365 | 94.2% | 92.9% | 92.9% | | Mealie | 310 | 94.2% | 94.2% | 94.2% | | n8n | 143 | 88.8% | 89.5% | 89.5% | | Shlink | 137 | 92.0% | 92.0% | 92.0% | | Sentry | 226 | 40.7% | 51.3% | 40.7% | **The extra sessions supplied the evidence; what they showed was a modelling gap each time, so agreement barely moved:** - **Mealie:** the capture's recipe create answers only a slug and is never read back, so all recipe plans were `no-observation`. The later-call rule now builds them from the training session. The create is then contradicted by a list that keeps its recorded members. - **Komga:** the training session lists after a create. That list keeps its recorded members (2 where Komga has 3), so the gate now withholds the create (`side-effect`) that the capture alone served. The drop is that create no longer being served. - **Lists:** membership is decided per tool, and each of these tools has filtered recordings, so their lists never show created entities. Per-line membership is the next fix. - **Shlink:** stores sent tags sorted, so its update fails the echo rule in every build. - **n8n:** the delete-before-archive 400 is not modelled. - **Sentry:** the gain is recency, not modelling. The time-bucketed views take their latest initial values from sessions recorded closer to the held-out reads. Joined reading loses those reads. Reading the same recordings as one session never did better than sessions. **Found along the way:** - A joined HAR is bounded by the HAR decoder's fixed 192 MiB heap: n8n's needed one 10.9 MB response (over the state's 256 KiB body limit) left out of the added session. - n8n's first training session carried an unrelated write from the capture's prefix and was re-recorded. Details are in `docs/specs/agent-environments/recording-sessions/verification.md`. **Remaining proof gaps:** - REA cannot check that sessions shared a starting state; the measurement used snapshot restores. - The held-out sessions cover each application's written collections and seeded reads, not every tool. For evidence-bearing changes: - [x] Observed, derived, and inferred claims remain distinguishable. - [x] Artifact identity, source provenance, and failed attempts remain preserved. - [x] Unsupported, incomplete, unavailable, or uncertain outcomes remain visible. ## Validation performed Run on macOS with Node v22.23.1 at e49361ca. - `npm run build:cached`, `npm run check:fast`, `npm run format:check`, `npx knip`, `npm run scan:todos`, `npm run docs:check`, `npm run verify:test-discovery`, `npm run mcp-catalog:check` — passed. - `npm run jscpd` — 3.46% over 3603 files. - `npx vitest run --project domain --project services` — 7484 passed, 5 pre-existing skips. - `npm run test:prepare`, then `npm run test:focused --` the environment serve and export CLI boundaries, the serve and served-writes MCP boundaries, the analysis-view CLI and MCP boundaries, the compact input schema boundary and tool schema validity — 8 files, 78 passed. ## Compatibility, safety, and release - Breaking changes or migration steps: none. Without `session_starts`, views and Evidence ids are unchanged. With the later-call rule, a plan whose first call could not be built may now be built and served from a later call. - Real Hopper/Ghidra, browser, or OS coverage: not applicable. - Package or release metadata impact: none. - Security, privacy, process, or containment review: no new inputs beyond ordinals; the verification's recordings were made without headers or cookies. ## Author readiness Human owner:  - [ ] I reviewed the final diff and relevant existing code, understand the change and its edge cases, and take responsibility for the complete PR, including AI-assisted work. - [ ] I checked the evidence above, confirmed it supports the current change, and recorded meaningful validation gaps. - [ ] I declare this PR ready for initial maintainer review. ## Before merge - [ ] This PR addresses a concrete problem or an agreed enhancement. - [ ] Material scope and authority questions are resolved, and any required maintainer scope decision covers the final diff. - [ ] The PR has one focused outcome and the title follows `type(scope): outcome`. - [ ] Related issue is linked, or the reason for not linking one is stated above. - [ ] Tests cover changed observable behavior and meaningful failure paths. - [ ] Owning docs, contracts, and generated metadata are updated where needed. - [ ] User-visible CLI/MCP changes include representative output. - [ ] I checked the final diff for secrets, unrelated cleanup, and unsupported claims. 🤖 Generated with [Claude Code](https://claude.com/claude-code) --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi

2026-11-01 21:44:53 +00:00

closed this issue
added the
pull-request
label

bsakaguchi referenced this issue

2026-11-03 13:38:59 +00:00

[PR #7037] [MERGED] fix(snapshot): decline ambiguous exact evidence replay #7594

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

morluto/rea#7037

No description provided.

Rows
Columns