simt-dev 0.7.1 → 0.8.0

package/CHANGELOG.md

@@ -4,7 +4,42 @@ All notable changes to this package are recorded here. The format follows
 [Keep a Changelog](https://keepachangelog.com/en/1.1.0/), and the project follows
 [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
-## [Unreleased]
+## [0.8.0] — 2026-11-01
+
+### Added
+
+- Every writer runs through one plan/apply engine: `init`, `sync`, `ci`, `ignore`, `rules decisions`, `pack new`, `upgrade`, `hooks install`, `agent-hooks install` and the agent plan derive create/update/remove from the bytes, journal one set per run, and a rehearsal reports the plan and writes nothing.
+- An append-only event log every store writes to (`.simt/local/events.jsonl`): journal commits and undos, gate runs, lease acquire/release and lesson captures all land in it; the ledger, lease and lessons files remain the projections readers already use.
+- Lease schema v2: a claim names a resource (a file, or a directory and its subtree), its holder, and a fencing token that increments per acquisition, with renew and release; a v1 lease file is read as path resources rather than dropped.
+- The MCP list methods answer one page and name the cursor for the next; a subscription request is refused with its reason, and a cancelled call is answered with its interruption. The surface stays a reader. — [ADR-126](docs/decisions/ADR-126-the-mcp-surface-stays-a-reader-so-there-is-no-write-path-to-build.md)
+- Index storage v4: generic nodes and edges with ownership, an `openIndex()` API, and O(n) PageRank; one `coverage()` and one `governs()` sit behind every reader that answers from the index.
+- The ledger gets a seeded private index, a tail read and a rotation.
+- Ecosystem/language data records — runner, gates, card, test conventions, extraction, resolution — ship in every channel beside the harness records. — [ADR-013](docs/decisions/ADR-013-genericity-by-facts-and-local-profiles.md)
+- The release process is a repo-local skill the tool recognises and injects when a release task is in play.
+- Bare `simt status` is the repository snapshot — the last check, git hook routing, which harnesses decide tool calls, evidence for the tree on disk, active leases, tracker counts and the last journal set — while `--line` keeps its cached one line. Each slot is read through the module that owns it and carries its own error, so one unreadable file does not blank the answer.
+- `simt explain` answers the five policy ids a denial names (`push`, `rewrite`, `hooks-disable`, `generated`, `lease`), so a refusal can say `run: simt explain push`.
+- A spec row declares `aliases`, and an alias resolves to that row before the registry is consulted: `skills list`, `skills lint`, `skills select`, `harness cost`, `harness hook`, `harness install`, `harness plugin`, `plan start`, `plan finish` and `plan abandon` answer, and help keeps listing the path.
+- The bench has one registry per seam — arms, adapters, task schema, records and graders — and a record written by an earlier version is read as written rather than migrated.
+- Production reads test conventions without the bench: `src/testing/` and `src/measure/` are separable from it.
+
+### Changed
+
+- `run(command, input)` answers with what the command produced — text, documents, notes, warnings, failure and exit code — instead of printing it, and one call carries one presentation state, not one per process.
+- The harness facet data — the surfaces each harness loads, its pointer files, its loader chain, its hook dialect and payload keys, its plugin components — moves out of the engine modules into `harnesses/`: one JSON record per harness, shipped in the tarball, the wheel and the sdist alongside the rule bodies. The views read the record by id, so a new harness is a record, not a table edit across four modules. — [ADR-013](docs/decisions/ADR-013-genericity-by-facts-and-local-profiles.md)
+- The git and process plumbing is one module: one runner behind identity, hooks, the store and `tracked()`, a bounded `spawnBounded`, and shared frontmatter and tokenizer helpers.
+- `agent-hooks install` folds into `init`, and help flags the prerequisites a command needs.
+- The plan tracker reads `docs/research`, and small work — fewer than three slices — skips the plan file.
+- The README's command list renders from the spec table, so it cannot drift from the commands the binary dispatches.
+- Bare `simt bench` summarises the vault (`bench report`) instead of mining into it: a reader who types a group name has asked a question, not given consent.
+- The projection gate reads the plan the writer built instead of rebuilding it, and gate planning lives in `src/gates/index.mjs`, so the manifest module keeps only its writer. — [ADR-013](docs/decisions/ADR-013-genericity-by-facts-and-local-profiles.md)
+- No source or test module is over 1,000 lines with one exclusion (the generated skill registry), and the ceiling's remaining long files — rules, MCP tools and schemas, the hooks tests, `init` — are split rather than excused.
+- The tolerated-flag window moves to 0.9.0. It named 0.8.0, and closing it here would refuse flags that work: the spec has not yet declared every flag some command honours (`ci --dry-run` among them). A window that closes over an incomplete declaration list turns working invocations into usage errors, so it closes when the list is complete.
+
+### Fixed
+
+- A v2 lease write keeps the flat aliases and caller fields a v1 reader expects.
+- The lesson event is emitted where the record is written, so it lands in the event log.
+- `simt status` no longer refuses a bare invocation: the answer it declined to give was five commands away.
 
 ## [0.7.1] — 2026-11-01
 

package/README.md

@@ -91,87 +91,176 @@ truncated or substituted download never becomes the command.
 
 ## What `simt` does
 
+<!-- simt:commands:begin -->
+<!-- Rendered from the spec by `simt docs readme --yes`; do not edit by hand. -->
+
 ```sh
-# Adopt, inspect, and maintain a repository
-simt init                           # adopt or scaffold; prints a plan; --yes writes
-simt doctor [--fix]                 # stack, facts, evidence, and config health (--fix applies safe remedies, journaled)
-simt check [--frozen]               # projection drift, manifest, lock, and declared checkers
-simt sync [--explain] --yes         # regenerate the projection
-simt upgrade --yes                  # move every consumer in the workspace to this SHA
-
-# Orient on a task and query the local code and governance graph
-simt brief "<task>"                 # one <= 2 KB packet: kind, commands, files, tests, ADRs, skills, lessons
-simt index [--embed]                # files, symbols, edges, FTS5 tokens (+ optional local embeddings)
-simt map --task "<task>"            # hybrid BM25 + PageRank + resolver file ranking (outline|names|signatures)
-simt impact <file|symbol>           # importers, covering tests, governance, and cross-repo consumers
-simt tests <path> [--cochange]      # tests covering a file via imports, naming, coverage, or co-change
-simt gate --changed                 # run only the tests the current diff reaches, with compacted output
-simt why [path]                     # ADRs, rules, skills, gates, and lessons governing a path
-
-# Rules, skills, decisions, and lessons
-simt rules list [--root <path>]     # rule catalogue and what applies to a repository
-simt rules show <id>                # read one rule body (including rules deferred in kernel mode)
-simt rules why <id|ADR-NNN> [--all] # the decision that settled a rule, and whether it still stands
-simt explain <CODE> [--json]      # what a check finding means, why it is reported, and how to fix it
-simt decide "<title>" --yes         # scaffold the next ADR with structured simt-affects / simt-supersedes
-simt lessons list                   # fail-to-fix, revert, review-block, and hook-denial candidates
-simt lessons review <id> --yes      # promote a candidate to docs/lessons/NNNN-<slug>.md
-simt skills index                   # tier-0 catalogue: every skill, one line each
-simt skills brief --task "<task>"   # tier-1 ranked skill selection within budget
-simt skills show <name>             # tier-2 skill body on demand (respects --max-bytes)
-simt skills why <name>              # the signal chain that selected or rejected a skill
-simt skills find <query>            # search skills by capability and keyword
-simt skills install <name> --yes    # pin a skill into the repository
-simt skills compile [<name>]        # compile to Agent Skills layout, split > 5,000-token bodies, report triggers
-simt skills eval [corpus]           # measure resolver precision and recall against a gold corpus
-simt graph lint [--spec agentskills]# verify the skill registry against skills on disk
-
-# Dual gates (git time and agent tool-call time) and execution evidence
-simt hooks install --yes            # install the shared POSIX git hook dispatcher
-simt hooks status                   # what each git event routes to, and whether it is wired
-simt hooks run <event>              # run pre-commit or pre-push gates now, as git would
-simt agent-hooks install --yes      # write pre/post-tool and prompt-submit hooks for active harnesses
-simt agent-hooks status             # what each harness will run, and whether Codex trusts it
-simt agent-hook <event> --harness   # evaluate pre-tool-use, post-tool-use, or prompt-submit
-simt evidence                       # gate and test runs recorded against the current git tree
-simt evidence check <msg-file>      # verify a commit message does not claim unrecorded passing runs
-simt pr-body [--base <ref>]         # a PR body: the evidence block for this tree, changed
-                                    # decisions, gate state, and residual risk in ADR-008 vocabulary
-
-# Plans, context metering, workspace learning, benchmarks, local models, and MCP
-simt plan status                    # count queued, in-progress, complete, and accepted tracker rows
-simt plan next [<id>]               # print the next ready slice section (~1 KB)
-simt plan done <id> [<sha>] --yes   # mark a tracker row complete with its commit SHA and test command
-simt status --line                  # one-line gate state from the last check, for prompts (< 50 ms)
-simt undo --yes                     # restore the last journaled set of writes, if untouched since
-simt slice start <id> --yes         # a worktree on branch slice/<id> with the brief preloaded; row In progress
-simt slice finish <id> --yes        # run the pre-commit gates, commit a Conventional subject from the
-                                    # row's title, and record the SHA and --test command on the row
-simt lease [<paths>] --session      # claim files for this session (--ttl, default 1h) or list claims;
-                                    # the pre-edit hook refuses another session's edit until the TTL
-simt tokens [--repos] [--session]   # measure bytes loaded at session start or split a transcript
-simt harness detect [--user]        # detect configured harnesses at repo and user level
-simt harness loaders [--table]      # what each harness loads, in what order, and where its cap truncates
-simt shared plan                    # rules repeated across a workspace and the bytes saved by sharing
-simt shared install --yes           # write shared rules once at user level and record the repo delta
-simt compare                        # workspace matrix, violated rules, noisy/slow gates, and ADR drafts
-simt local doctor [--allow-remote]  # probe configured OpenAI-compatible local chat and embed endpoints
-simt bench mine --yes               # mine verified fail-to-pass tasks from git history into the vault
-simt bench run --yes                # evaluate arms (bare, current, kernel, brief) in isolated exports
-simt bench report [--compare a b]   # resolved rate, quality, cost, CPI, bootstrap CIs, McNemar/Wilcoxon
-simt bench ablate --yes             # leave-one-out rule ablation ranked by value per KB
-simt bench probes --yes             # 12 deterministic rule-compliance probes
-simt bench variants --yes           # evaluate simt-variants slots and lock winners in simt.lock.json
-simt review [--local]               # an advisory ADR-008 verdict skeleton over the staged diff
-simt distill <skill|doc>            # a budgeted proposal summary with line citations, for a person to merge
-simt pack new <name> --source <spec> # scaffold a versioned pack: pack.json, bench.json, one rule
-simt pack lint [dir] [--require-measured] # validate a pack and state its measured ΔR
-simt pack publish [dir] --yes       # publish an npm-sourced pack (git packs name the tag-and-push)
-simt mcp [--list]                   # serve the 15 read-only tools over stdio JSON-RPC
-simt completions <bash|zsh|fish>    # a completion script generated from the command registry
-simt ignore [--yes] [--diff]        # render the one .simtignore list into .cursorignore,
-                                    # .aiderignore, .geminiignore and Claude's permissions.deny
+# agent-hook
+simt agent-hook <event> --harness            # the harness entry point: decide one tool call, or deliver the prompt packet
+
+# agent-hooks
+simt agent-hooks install --yes               # write this repository's policies into each harness hooks file
+simt agent-hooks status                      # what each harness will run, and what is not trusted yet
+
+# agents
+simt agents install --yes                    # write the Simt-Agent into each harness agents file
+simt agents status                           # what each harness carries, and what is not installed yet
+
+# bench
+simt bench ablate --yes                      # measure one arm of the brief against its ablation
+simt bench critic                            # score the review critic against the labelled bench set
+simt bench export --yes                      # write the sealed bench corpus as one signed document
+simt bench import                            # verify a sealed corpus and load it into the vault
+simt bench mine --yes                        # mine tasks out of a repository history
+simt bench probes --yes                      # run the deterministic probes against a workspace
+simt bench report [--compare a b]            # summarise the runs recorded in the vault
+simt bench run --yes                         # run agents against the mined tasks
+simt bench variants --yes                    # run one task across its variant prompts
+
+# brief
+simt brief "<task>"                          # the packet for a task: rules, skills and the files it names, within a budget
+
+# check
+simt check                                   # report stale or missing generated regions and every check finding
+
+# ci
+simt ci                                      # refresh simt-dev's own .github workflow and Dependabot files
+
+# compare
+simt compare                                 # the adoption and drift matrix across every sibling repository
+
+# completions
+simt completions <bash|zsh|fish>             # print a bash, zsh or fish completion script generated from this spec
+
+# decide
+simt decide "<title>" --yes                  # scaffold a numbered decision record with its frontmatter
+
+# distill
+simt distill <skill|doc>                     # turn a repository into a skill or rule proposal
+
+# docs
+simt docs readme                             # render the README command list from the spec
+
+# doctor
+simt doctor [--fix]                          # report what is wrong with this installation
+
+# evidence
+simt evidence                                # the runs recorded against the tree on disk
+simt evidence check <msg-file>               # a commit message claiming a run this tree does not have
+
+# explain
+simt explain <CODE>                          # what one check finding code means, and how to fix it
+
+# gate
+simt gate --changed                          # the tests a change reaches, and only those
+
+# graph
+simt graph lint [--spec agentskills]         # check the skill registry against the skills on disk
+
+# harness
+simt harness detect [--user]                 # which harness is installed, and what it carries
+simt harness loaders [--table]               # the instruction files each harness loads, and what they cost
+
+# hooks
+simt hooks install --yes                     # install the shared dispatcher and report what it would run
+simt hooks run <event>                       # run one git event's gates now, as git would
+simt hooks status                            # what each git event routes to, and whether it is wired
+
+# ignore
+simt ignore [--yes] [--diff]                 # render the one ignore list into every harness ignore file
+
+# impact
+simt impact <file|symbol>                    # who a change reaches: importers, their tests and the contract
+
+# index
+simt index [--embed]                         # rebuild the local file index from git blob hashes
+
+# init
+simt init [path]                             # adopt this repository: manifest, region, hooks, agents and agent-hooks
+
+# lease
+simt lease [<paths>] --session               # claim files for one session, or list the claims
+
+# lessons
+simt lessons capture                         # scan the ledger and history for new lesson candidates
+simt lessons list                            # deterministic lesson candidates and the promoted lessons
+simt lessons review <id> --yes               # promote one lesson candidate into docs/lessons
+
+# local
+simt local doctor [--allow-remote]           # the local model endpoints, and whether they answer
+
+# map
+simt map --task "<task>"                     # the files a task is most likely to touch, within a byte budget
+
+# mcp
+simt mcp [--list]                            # serve the read-only MCP surface over stdio
+
+# pack
+simt pack lint [dir] [--require-measured]    # lint a rule pack in the working directory
+simt pack new <name> --source <spec>         # scaffold a rule pack
+simt pack publish [dir] --yes                # publish a rule pack
+
+# plan
+simt plan done <id> [<sha>] --yes            # mark one tracker row complete, with its SHA and its evidence
+simt plan next [<id>]                        # one ready slice's section, or the one named
+simt plan status                             # count the tracker rows
+
+# plugin
+simt plugin <claude|gemini|cursor|opencode>  # generate a harness package from the same tables the hooks use
+
+# pr-body
+simt pr-body [--base <ref>]                  # the pull request body: evidence, decisions, gate state and residual risk
+
+# review
+simt review [--local]                        # review the staged diff against the estate rules
+
+# rules
+simt rules decisions                         # every known decision, and which layer each came from
+simt rules list [--root <path>]              # the rule catalogue, marked with what applies to this repository
+simt rules show <id>                         # one rule body
+simt rules why <id|ADR-NNN> [--all]          # the decision that settled a rule, and whether it is still in force
+
+# shared
+simt shared install --yes                    # install the shared rules into a reader home directory
+simt shared plan                             # what the shared install would write, and where
+
+# skills
+simt skills brief --task "<task>"            # the packet for a task, read from the skill registry
+simt skills compile [<name>]                 # compile a skill to the Agent Skills standard
+simt skills eval [corpus]                    # score the skill registry against transcripts
+simt skills find <query>                     # skills that match a query
+simt skills index                            # rebuild the skill index
+simt skills install <name> --yes             # install skills into the workspace
+simt skills show <name>                      # one skill packet
+simt skills why <name>                       # which skill covers a task
+
+# slice
+simt slice abandon <id>                      # remove a slice worktree, delete its branch, and return the row
+simt slice finish <id> --yes                 # run the gates, commit the slice and record its evidence
+simt slice start <id> --yes  |  -- "<text>"  # give a slice its worktree, branch and brief; -- starts one from a prompt, no row
+
+# status
+simt status [--line]                         # the repository snapshot, or one line of the last check
+
+# sync
+simt sync [--explain] --yes                  # re-render the generated regions from the rule content
+
+# tests
+simt tests <path> [--cochange]               # the tests that exercise a file
+
+# tokens
+simt tokens [--repos] [--session]            # count tokens per instruction file and per skill
+
+# undo
+simt undo --yes                              # restore the last journaled set of writes, if untouched
+
+# upgrade
+simt upgrade --yes                           # replace this installation with the pinned release
+
+# why
+simt why [path]                              # what governs a file: decisions, rules, skills, gates and lessons
 ```
+<!-- simt:commands:end -->
 
 `init`, `sync`, `upgrade`, `hooks install`, `agent-hooks install`, `shared install`,
 `skills install`, `plan done`, `lessons review`, `decide`, `ignore`, `pack new`,
@@ -239,7 +328,7 @@ on top of `0.6.0`:
   for parallel agents, and `simt pr-body` assembled from the evidence ledger.
 
 The last two rows of review `0002` (`S38` workspace token cost, `S39` estate policy
-as a pack) are closed by ruling in ADR-102: the instrument, the declaration and the
+as a pack) are closed by ruling: the instrument, the declaration and the
 non-blocking finding shipped, the blocking gate and the registry-carrying pack were
 refused by ruling, and the estate-side move is recorded as whose commit it waits on.
 Everything in plans `0001`–`0003` and reviews `0001`–`0002` is Complete or Accepted.

package/docs/decisions/ADR-014-workspace-is-configuration.md

@@ -133,7 +133,7 @@ a feature; `rules why` is the provenance mechanism the whole one-canonical-home
 on, and it cannot depend on a git checkout being present.
 
 `docs/plans` ships with it, because every non-zero exit prints a remedy pointing at
-`docs/plans/0002-simt-dev-cli.md section 11`.
+`docs/research/0002-simt-dev-cli.md section 11`.
 
 ### 6. "Absent", "wrong" and "moved" are three different facts
 

package/docs/decisions/ADR-025-a-baseline-is-a-region-in-a-file-the-repo-owns.md

@@ -4,7 +4,7 @@ Status: accepted · 2026-10-06 · rulings 1 and 2 narrowed by ADR-026 · ruling
 **Affects:** `src/projection/baselines.mjs`, `src/projection/markers.mjs`,
 `src/projection/project.mjs`, `src/commands/init.mjs`, `test/baselines.test.mjs`,
 `test/leaks.mjs`, `test/surfaces.test.mjs`. Settles §3.1 row 5 and row 7 of
-`docs/plans/0001-centralise-agent-rules.md` for the files it names, and leaves the rest ruled
+`docs/research/0001-centralise-agent-rules.md` for the files it names, and leaves the rest ruled
 out by name rather than outstanding.
 
 Rulings 1 and 2 did not survive whole: ADR-026 rules that a config whose grammar cannot

package/docs/decisions/ADR-026-a-config-without-a-region-is-a-file-the-repo-owns.md

@@ -5,7 +5,7 @@ Status: accepted · 2026-10-06
 `src/commands/help.mjs`, `src/projection/project.mjs`, `test/ci.test.mjs`.
 **Narrows:** ADR-025 rulings 1 and 2. **Extends:** ADR-023 (bytes copied, written once),
 ADR-024 (simt-dev's own copies are refreshed by `simt-dev ci`). Settles the Dependabot
-item §3.1 row 8 of `docs/plans/0001-centralise-agent-rules.md` named as still missing.
+item §3.1 row 8 of `docs/research/0001-centralise-agent-rules.md` named as still missing.
 
 ## Context
 

package/docs/decisions/ADR-027-the-one-gitattributes-line-that-is-neither-dangerous-nor-empty.md

@@ -5,7 +5,7 @@ Status: accepted · 2026-10-06
 `test/baselines.test.mjs`, `test/surfaces.test.mjs`.
 **Supersedes:** ADR-025 ruling 9, which ruled `.gitattributes` out for now and required this
 record. Settles the `.gitattributes` item in §3.1 row 5 of
-`docs/plans/0001-centralise-agent-rules.md`.
+`docs/research/0001-centralise-agent-rules.md`.
 
 ## Context
 

package/docs/decisions/ADR-028-a-base-file-is-a-floor-so-the-disagreement-it-cannot-repair-becomes-a-check.md

@@ -5,7 +5,7 @@ Status: accepted · 2026-10-07
 `test/tsconfig.test.mjs`. **Narrows:** ADR-026 ruling 2 (ownership does not depend on
 content being mergeable) by applying it to a file whose mechanism is `extends` rather than
 a copied byte range. Settles the `tsconfig.base.json` item of §3.1 row 5 and row 7 of
-`docs/plans/0001-centralise-agent-rules.md`.
+`docs/research/0001-centralise-agent-rules.md`.
 
 ## Context
 

package/docs/decisions/ADR-029-the-pins-a-repository-declares-are-its-own.md

@@ -3,7 +3,7 @@
 Status: accepted · 2026-10-07
 **Affects:** `src/toolchain.mjs`, `src/projection/project.mjs`, `test/toolchain.test.mjs`.
 **Settles:** the `packageManager`/`engines`/`.nvmrc` items of §3.1 row 5 of
-`docs/plans/0001-centralise-agent-rules.md`. **Extends:** ADR-026 ruling 9 (absence and
+`docs/research/0001-centralise-agent-rules.md`. **Extends:** ADR-026 ruling 9 (absence and
 disagreement are reportable; content is not) to a file this tool authors no part of.
 
 ## Context

package/docs/decisions/ADR-030-what-a-plan-does-not-ship-is-a-decision-with-a-record.md

@@ -1,7 +1,7 @@
 # ADR-030 — what a plan does not ship is a decision with a record, not a silence
 
 Status: accepted · 2026-10-07
-**Affects:** `docs/plans/0001-centralise-agent-rules.md` §3.1 row 5 and row 7. Ships no
+**Affects:** `docs/research/0001-centralise-agent-rules.md` §3.1 row 5 and row 7. Ships no
 code. **Refuses:** the commitlint baseline and the shared ESLint block named in row 5 and
 row 7, and records why in the plan itself rather than leaving the rows to imply that two
 mechanisms are merely unwritten.

package/docs/decisions/ADR-048-a-baseline-carries-only-keys-a-reader-honours.md

@@ -2,7 +2,7 @@
 
 Status: accepted · 2026-10-09 · amends ADR-025 (withdraws ruling 8, corrects its context)
 **Affects:** `src/projection/baselines.mjs`, `test/baselines.test.mjs`,
-`docs/plans/0001-centralise-agent-rules.md`. Closes S30 of
+`docs/research/0001-centralise-agent-rules.md`. Closes S30 of
 `docs/reviews/0002-full-review-adoption-and-public-release.md`.
 
 ## Context

package/docs/decisions/ADR-082-the-tracker-override-is-a-manifest-key.md

@@ -1,6 +1,6 @@
 # ADR-082 — The tracker override is a manifest key, and init keeps what it did not write
 
-Status: **accepted** · 2026-10-16 · supersedes ADR-081 rulings 1 and 2
+Status: **accepted** · 2026-10-16 · supersedes ADR-081 rulings 1 and 2 · ruling 1 superseded by ADR-128
 
 ## Context
 

package/docs/decisions/ADR-126-the-mcp-surface-stays-a-reader-so-there-is-no-write-path-to-build.md

@@ -0,0 +1,68 @@
+---
+simt-affects: ["src/mcp/server.mjs", "src/mcp/tools.mjs", "docs/reviews/0003-pre-plan-0004-review.md"]
+simt-supersedes: []
+---
+
+# ADR-126 — the MCP surface stays a reader, so there is no write path to build
+
+Status: **accepted** · 2026-11-01
+
+## Context
+
+Review 0003's Q66 lists "the two-step write path" among the clauses still open for the
+MCP surface. The clause reads as work to do; the file says otherwise. `src/mcp/server.mjs`
+opens with three decisions, the third of which is that no mutating tool is exposed —
+`init`, `sync`, `hooks install`, `skills install` and `upgrade` are absent on purpose,
+because an editor that can call a tool can call it without a human reading a diff first.
+
+Those two statements have stood in the same repository for months without meeting. The
+question the row asks and the rule the file states are the same question, and the answer
+was written down where the row's reader would not look.
+
+The argument for exposure is real and worth stating: a client that can read a repository
+can also want to fix it, and a plan-then-apply pair would let it see what would change
+before changing anything. What it would not do is make the change reviewed. `--yes` on
+the command line is a human decision with a preview in front of it; the same flag arriving
+inside a tool call is a model deciding to pass it.
+
+## Ruling
+
+### 1. No tool on this surface writes
+
+Every tool returns a value and changes nothing. The capability block advertises no write
+capability, and this is not a gap to be filled later: a tool that writes is a different
+surface with a different review obligation, and it is not this one.
+
+### 2. A caller that wants to write runs the command
+
+The commands already do plan-then-apply properly: they print a preview, refuse without
+`--yes`, journal what they write, and `undo` restores it. A client that wants that power
+runs `simt <command>` as a subprocess, where the human who opened the editor sees the
+preview and passes the flag — or does not.
+
+### 3. Exposing a writer needs a superseding record, not a patch
+
+If a future surface must write, it arrives with its own decision: what the plan token
+proves, who approves the apply, how a half-applied run is undone, and what the journal
+says. Approving a diff in an editor is a claim the server cannot verify, and a
+confirmation prompt is a user interface the server does not own.
+
+## Consequences
+
+- Q66's write-path clause closes as withdrawn, not deferred. The row names this record so
+  the next reader does not reopen it as unfinished work.
+- The read-only guarantee in the server's header, the capability block and the tests now
+  have one record behind them instead of three comments.
+- A tool that returns `{ written: false }` for every input is not a compromise: it is the
+  shape that makes a client believe a write happened.
+
+## Rejected
+
+- **A plan/apply pair behind a confirmation prompt.** The prompt is rendered by the client,
+  which may not render one, and the server cannot tell an approved apply from an unapproved
+  one. A safety property that depends on another program's UI is not a property.
+- **A write tool gated on `--yes` passed through the arguments.** The flag would be asserted
+  by the same model that decided to write; the preview would be read by nobody.
+- **Exposing `sync` alone because it is idempotent.** It rewrites the generated region of
+  every repository in a workspace. Idempotence says a second run changes nothing, not that
+  the first run was reviewed.

package/docs/decisions/ADR-127-the-mcp-tool-list-is-written-by-hand-and-roots-is-not-asked-for.md

@@ -0,0 +1,65 @@
+---
+simt-affects: ["src/mcp/tools.mjs", "src/commands/spec.mjs", "test/mcp-spec.test.mjs", "docs/reviews/0003-pre-plan-0004-review.md"]
+simt-supersedes: []
+---
+
+# ADR-127 — the MCP tool list is written by hand and held to the spec, and roots is not asked for
+
+Status: **accepted** · 2026-11-01
+
+## Context
+
+Two clauses of review 0003's Q66 read as work still to do: "`tools/list` generated from
+spec rows with `mcp` set", and "`roots`". Both were written before the spec rows existed,
+when the four lists this review found disagreeing were the live problem and generation
+looked like the answer to all of them.
+
+The spec rows now exist, and they carry `path`, `summary`, `flags` and, for the commands
+the surface exposes, `mcp: {name, flags}`. That is enough to say *which* flags a tool
+exposes and not enough to say what they are: no types, no per-argument descriptions, and
+no vocabulary shared with the transport. `check` declares no `mcp` flags and its tool
+takes `root`; `compare` declares none and its tool takes `root`, `limit`, `cursor` and
+`concurrency`. An `inputSchema` derived from the rows would describe `--dry-run` as a
+property and omit `root`, which is a contract a client cannot call.
+
+`roots` is the other direction. It is a client capability: the server sends `roots/list`
+and the client answers with the directories it may touch. `src/mcp/server.mjs` answers
+requests and never sends them — the only thing it writes unprompted is a notification for
+a progress token the client supplied. A server that asks a question needs a correlation
+table for replies, a timeout for a client that never answers, and a fallback for the
+clients that do not implement the capability; and every tool here already takes its root
+as an argument (`root`, `workspace`, or `--paths`), which is the same information stated
+once, by the caller, where it can be read in the call rather than inferred from a
+handshake.
+
+## Ruling
+
+### 1. The tool list stays written by hand, and tests hold it to the spec
+
+The spec row is the source of the *command*: its path, its summary and its flags. It is
+not the source of the transport's schema, which carries descriptions, types and defaults
+the row does not have, and arguments the command line does not spell as flags.
+
+What generation was meant to prevent — a tool with no command, a command nobody can reach
+over the transport, an advertised flag the tool refuses — is enforced in
+`test/mcp-spec.test.mjs`, in all three directions, plus `MCP_ONLY` for the one tool that is
+deliberately not a command (`federation_query`, which names its reason). A hand-written
+list under a test that fails on every kind of drift is the same guarantee as a generated
+one, without a schema nobody would trust enough to keep descriptions in.
+
+### 2. `roots` is withdrawn, not deferred
+
+No tool is changed to accept an implicit root, and the server gains no request path. A
+client that wants a tool to run against a directory passes it as the argument the schema
+already names. If a future capability genuinely needs an answer from the client, that is
+a different surface with its own correlation and timeout design, and it needs its own
+record.
+
+## Consequences
+
+- Reading `mcp: {name, flags}` tells a reader which flags reach the transport; reading
+  `src/mcp/tools.mjs` tells them what a caller sees. Both are checked against each other.
+- Adding a second MCP-only tool means a `MCP_ONLY` entry with a reason, which is a place a
+  reviewer looks, rather than a hole in the table.
+- The 2 KB `tools/list` budget keeps being spent on descriptions a reader can act on,
+  because nothing regenerates them from a summary phrase.

package/docs/decisions/ADR-128-a-research-plan-is-a-plan-and-the-tracker-reads-it.md

@@ -0,0 +1,42 @@
+---
+simt-affects: ["templates/plan/tracker.json", "src/plan/index.mjs", "src/projection/manifest.mjs", "test/rules.test.mjs", "test/plan.test.mjs"]
+simt-supersedes: ["ADR-082"]
+---
+
+# ADR-128 — A research plan is a plan, and the tracker reads it
+
+Status: **accepted** · 2026-11-01 · supersedes ADR-082 ruling 1
+
+## Context
+
+The four founding plans left `docs/plans` for `docs/research`: completed work of a
+different kind — research — but plans nonetheless, with tracker tables, slice ids and wave
+dependencies. The tracker reads only the directories its template lists, so the move
+de-indexed all four at once. A wave dependency that resolved through them broke with it:
+`R01` in the research plan 0003 depends on `review 0002 Wave 0`, and the slices it names
+live in the de-indexed documents.
+
+The directory set was never code. ADR-081 and ADR-082 settled the mechanism — a template
+lists any number of directories, each with a `kind`, a `word` and a `path`, and the parser,
+the help text and `plan status` all read it — and ADR-082 ruling 1 fixed the *delivered*
+set at two. Fixing the delivered set is a ruling, and this is the new one.
+
+## Ruling
+
+1. **The delivered template lists `docs/research` as a third directory, kind `research`,
+   word `research`.** A repository that sets nothing gets all three; a repository that
+   overrides `plans` replaces the `directories` section wholesale, exactly as before, so no
+   existing override is disturbed.
+2. **The directory is data, not a special case.** Nothing in `src/` names `research`; the
+   parser, help and status keep reading the template. A fourth directory arrives the same
+   way: one entry in the file, one record here.
+
+## Consequences
+
+- `simt plan status` names three directories, and the research plans are indexed again:
+  their slices count, and a dependency may name one as `research 0003 R01`.
+- Template order is index order. `research` sits between `plans` and `reviews`, which keeps
+  the moved documents ahead of the reviews they depend on — the same relative order they
+  had while they were in `docs/plans`.
+- The tests that walk the tracker's documents (`test/rules.test.mjs`) read the template's
+  directories instead of a hardcoded pair, so a future entry reaches them without an edit.

package/ecosystems/bun.json

@@ -0,0 +1,6 @@
+{
+  "id": "bun",
+  "dependabot": [
+    { "order": 20, "fact": "has-bun-lock", "value": "bun", "manifest": "bun.lock" }
+  ]
+}

package/ecosystems/cargo.json

@@ -0,0 +1,25 @@
+{
+  "id": "cargo",
+  "facts": ["has-cargo-toml"],
+  "tool": "cargo",
+  "manifestOrder": 1,
+  "gates": {
+    "preCommit": ["format", "lint"],
+    "prePush": ["test"],
+    "commands": {
+      "format": "cargo fmt --check",
+      "lint": "cargo clippy -- -D warnings",
+      "test": "cargo test"
+    }
+  },
+  "card": {
+    "build": { "order": 1, "condition": { "fact": "has-cargo-toml" }, "command": "cargo build" },
+    "focused": [
+      { "order": 5, "kind": "Rust", "condition": { "fact": "has-cargo-toml" }, "template": "cargo test {name}" }
+    ]
+  },
+  "dependabot": [
+    { "order": 6, "fact": "has-cargo-toml", "value": "cargo", "manifest": "Cargo.toml" }
+  ],
+  "languages": [{ "lang": "rust", "exts": [".rs"] }]
+}