loadout-ai 0.9.0 → 0.9.2

package/CHANGELOG.md

@@ -1,5 +1,102 @@
 # Changelog
 
+## Unreleased
+
+## 0.9.2 - 2026-09-06
+
+### Changed
+
+- Coordination discussion output now removes duplicate lifecycle noise, shows a
+  single outcome section, and identifies the provider in every provider error.
+- Maximum Library preparation now skips packages that cannot be safely prepared
+  and clearly reports that the remaining reviewed library can still install.
+
+### Fixed
+
+- Bound overlong provider-synthesized decisions to the coordination event limit
+  instead of failing after all paid discussion turns have completed.
+- Materialize multi-path ownership claims once per normalized path so snapshots
+  and status output do not repeat the full claim for every path.
+- Close the Claude Code subprocess stdin explicitly, avoiding its empty-input
+  wait before a bounded coordination turn begins.
+- Forward cancellation and timeout signals natively to both Claude Code and the
+  Codex SDK so a timed-out paid turn cannot keep running while Loadout starts a
+  replacement turn.
+- Restore files before reapplying read-only directory modes and reject snapshot
+  modes containing special permission bits.
+- Preserve strictly increasing coordination sequence order after checkpoint
+  compaction.
+- Treat all reviewed commits for a shared catalog repository as valid when
+  detecting upstream catalog drift.
+- Apply the default repository timeout to generic HTTPS and SSH Git sources as
+  well as catalog GitHub sources.
+
+### Security
+
+- Expand deterministic instruction-injection signatures and document the
+  repository-grounded threat model in `loadout-threat-model.md`.
+
+## 0.9.1 - 2026-09-05
+
+### Added
+
+- Snapshot capture and restore preserve Unix file permission modes, with
+  validation that rejects out-of-range values and tolerance for legacy
+  snapshots missing the `mode` field.
+- Provider cancellation via `AbortSignal` on adapter `start()` and
+  `submitTurn()` methods. Cancelled turns release the busy flag so new
+  turns can proceed.
+- Opt-in `loadout handoff ... --bundle <paths...>` context snapshots give the
+  receiving agent exact source context, with versioned references and visible
+  truncation metadata, while preserving all existing handoff messages.
+- Evidence-backed handoff completion with human criteria, optional shell-free
+  executable/JSON argv checks, bounded pass/fail evidence, and failed checks
+  that remain pending instead of being reported done.
+- Reusable built-in and project-local handoff templates with positional
+  placeholder expansion, bounded context bundles, and verification defaults.
+- Preview-first coordination setup, exact cross-boundary contract detection,
+  Git-author ownership mappings, and idempotent discussion-to-implementation
+  plans linked to their generated handoff tasks.
+
+### Changed
+
+- Log compaction preserves ownership, contract, and decision state by
+  extracting the latest checkpoint for each state key before removing events.
+  Double compaction is stable.
+- Auto-detected contracts handle mixed `import { A, type B }` syntax and
+  reject incomplete multi-line `const` declarations as unpublishable instead
+  of emitting truncated contract bodies.
+- Auto-detected contracts preserve exact supported declarations, expose a
+  source hash, distinguish current from stale revisions, and require both
+  `--publish` and `--yes`. Unsupported declarations require manual completion.
+- Git ownership confidence includes every author touching a directory and maps
+  agent identities explicitly with `agent=Git Author`.
+- Verification processes run outside the handoff log lock, so other agents can
+  continue sending tasks while a long check runs; completion remains atomic.
+
+### Fixed
+
+- Coordination lock recovery uses atomic rename so concurrent recoverers
+  cannot both delete and recreate the lock file. Post-acquisition token
+  verification ensures the winner still holds the lock after write.
+- Normalized contract-scanner paths across Windows and POSIX platforms and made
+  the verification cwd test independent of Windows short-path spelling.
+- Removed generated directories and redundant parent/child claims from
+  one-command coordination setup.
+
+### Security
+
+- Handoff bundles are limited to 20 project-relative text files, 32 KiB per
+  file and 50 KiB total. Creation and reading reject symlinks, traversal,
+  binary files, and internal `.git/` or `.handoff/` sources; common secrets are
+  redacted before owner-only atomic persistence.
+- Handoff text and completion output are redacted at the canonical write
+  boundary. Verification output is capped at 8 KiB per stream and command
+  execution requires `--run-verification`, is timeout-bounded, project-scoped,
+  and shell-free.
+- Template names are validated before every read or delete so traversal cannot
+  escape `.handoff/templates/`.
+
 ## 0.9.0 - 2026-09-03
 
 ### Removed

package/README.md

@@ -86,7 +86,7 @@ The second command detects your agents and previews the 30-skill Stable loadout.
 Nothing changes until you approve it. If anything goes wrong, start with the
 [user test guide](./docs/USER_TEST_GUIDE.md).
 
-For a reproducible install, pin the release: `npm install --global loadout-ai@0.9.0`.
+For a reproducible install, pin the release: `npm install --global loadout-ai@0.9.2`.
 
 ## Use it from inside your agent
 
@@ -121,13 +121,13 @@ If you pay for both Claude and a ChatGPT plan, the two agents cannot see each
 other. Loadout gives them a shared, append-only task log:
 
 ```bash
-loadout handoff codex "write unit tests for auth" --context "see src/auth.ts"
+loadout handoff codex "write unit tests for auth" --bundle src/auth.ts src/types.ts --verify "tests pass" --verify-command npm --verify-args '["test"]'
 ```
 
-That is the whole thing. The first send creates the log and adds a short managed
-block to `CLAUDE.md` and `AGENTS.md` telling each agent to check its inbox at the
-start of a session. Only the text between the `loadout:handoff` markers is
-managed; the rest of your file is left alone.
+Bundles snapshot exact, secret-redacted text (50 KiB total). Verification runs
+only with `--done --run-verification`, without a shell; failures stay pending. The first send adds
+managed inbox blocks to `CLAUDE.md` and `AGENTS.md`. Treat bundles as untrusted
+project data, review before committing, and never include credential files.
 
 ```bash
 loadout handoff codex      # what is waiting for codex
@@ -142,29 +142,20 @@ acknowledgements over a shared project event stream — not a merged context
 window. See the [live coordination guide](./docs/LIVE_COLLABORATION.md).
 
 ```bash
-loadout coord own claude-code src/api
-loadout coord contract checkout-api --agent claude-code \
-  --body "POST /api/checkout -> 201 { id: string }"
-loadout coord snapshot codex            # what codex needs to know
-loadout coord replay                    # full timeline as a story
+loadout coord start --agents claude-code,codex          # preview ownership
+loadout coord start --agents claude-code,codex --yes    # apply it
+loadout coord detect                                    # shared interfaces
+loadout handoff codex src/auth.ts --template write-tests
+loadout coord discuss start "REST or GraphQL?" --agents claude-code,codex --rounds 2 --max-turns 5
+loadout coord discuss implement <thread-id>             # preview linked tasks
 ```
 
-`loadout serve` exposes the same operations as MCP tools for both hosts.
-The optional provider bridge (`loadout coord agents bridge`) resumes sessions
-and delivers events at safe turn boundaries — never mid-turn.
-
-Start a bounded design discussion to have both agents challenge an approach
-before either writes code. Two rounds plus synthesis = five provider turns:
-
-```bash
-loadout coord discuss start "REST or GraphQL for checkout?" \
-  --agents claude-code,codex --rounds 2 --max-turns 5
-```
-
-Each response is explicitly public, linked to the previous response, and saved
-in the project audit trail. The discussion prompt forbids edits and tool use;
-review the final decision, then choose whether to implement it. Existing
-sessions work with `--sessions claude-code:<id> codex:<id>`.
+Contract publication and bounded design discussion implementation require a second `--yes`
+approval. Exact supported declarations are marked current or stale; ambiguous
+ones require manual review. The optional MCP server and provider bridge deliver
+events at safe turn boundaries, never during an active turn. The
+[coordination guide](./docs/LIVE_COLLABORATION.md) covers templates, Git-author
+mappings, the local dashboard, limitations, and the complete workflow.
 
 ## How it works
 
@@ -315,7 +306,7 @@ Catalog maturity: **53 sourced**, **53 technically inspected**, and **4 selected
 
 <!-- loadout:daily-discovery:start -->
 
-**Discovery snapshot (generated 2026-09-04):** [239 repositories observed](./docs/DISCOVERED.md), including 222 uncataloged review candidates and 17 repositories already in the inspected catalog.
+**Discovery snapshot (generated 2026-09-06):** [237 repositories observed](./docs/DISCOVERED.md), including 221 uncataloged review candidates and 16 repositories already in the inspected catalog.
 <!-- loadout:daily-discovery:end -->
 
 ## Trust and limits
@@ -353,27 +344,31 @@ Configured CI platforms describe a manually triggered workflow, not evidence tha
 
 ## Command reference
 
-| What it does                            | Command                                        |
-| --------------------------------------- | ---------------------------------------------- |
-| Beginner-friendly guided path           | `loadout guide`                                |
-| Preview the 30-skill Stable setup       | `loadout setup --mode stable`                  |
-| Apply after reviewing the preview       | `loadout setup --mode stable --yes`            |
-| Show managed packages and active skills | `loadout status` · `loadout library`           |
-| What fits this repository               | `loadout recommend --project .`                |
-| Project-specific active set             | `loadout optimize --project . --limit 30`      |
-| Scan existing skills across agents      | `loadout scan`                                 |
-| Check for source updates                | `loadout update`                               |
-| Find newly launched candidates          | `loadout discover --source all --queue`        |
-| Install Loadout's own skill             | `loadout skills install loadout-handoff --yes` |
-| Send a task to another agent            | `loadout handoff codex "write tests"`          |
-| Inspect shared agent state              | `loadout coord snapshot codex`                 |
-| Detect live provider runtimes           | `loadout coord agents detect`                  |
-| Debate one design with both providers   | `loadout coord discuss start "<topic>" ...`    |
-| Start the coordination MCP server       | `loadout serve`                                |
-| Agent health check                      | `loadout doctor`                               |
-| Rollback the latest managed change      | `loadout rollback`                             |
-| Preview complete removal                | `loadout uninstall`                            |
-| Full CLI reference                      | `loadout --help` · `loadout advanced`          |
+| What it does                            | Command                                                    |
+| --------------------------------------- | ---------------------------------------------------------- |
+| Beginner-friendly guided path           | `loadout guide`                                            |
+| Preview the 30-skill Stable setup       | `loadout setup --mode stable`                              |
+| Apply after reviewing the preview       | `loadout setup --mode stable --yes`                        |
+| Show managed packages and active skills | `loadout status` · `loadout library`                       |
+| What fits this repository               | `loadout recommend --project .`                            |
+| Project-specific active set             | `loadout optimize --project . --limit 30`                  |
+| Scan existing skills across agents      | `loadout scan`                                             |
+| Check for source updates                | `loadout update`                                           |
+| Find newly launched candidates          | `loadout discover --source all --queue`                    |
+| Install Loadout's own skill             | `loadout skills install loadout-handoff --yes`             |
+| Send a task to another agent            | `loadout handoff codex "write tests"`                      |
+| Use a reusable handoff template         | `loadout handoff codex src/auth.ts --template write-tests` |
+| Preview two-agent ownership setup       | `loadout coord start --agents claude-code,codex`           |
+| Detect shared contract candidates       | `loadout coord detect`                                     |
+| Inspect shared agent state              | `loadout coord snapshot codex`                             |
+| Detect live provider runtimes           | `loadout coord agents detect`                              |
+| Debate one design with both providers   | `loadout coord discuss start "<topic>" ...`                |
+| Turn a decision into linked tasks       | `loadout coord discuss implement <thread-id>`              |
+| Start the coordination MCP server       | `loadout serve`                                            |
+| Agent health check                      | `loadout doctor`                                           |
+| Rollback the latest managed change      | `loadout rollback`                                         |
+| Preview complete removal                | `loadout uninstall`                                        |
+| Full CLI reference                      | `loadout --help` · `loadout advanced`                      |
 
 Most mutating commands are dry runs first. Add `--yes` to apply.
 
@@ -396,7 +391,7 @@ npm run verify:full
 
 <!-- loadout:verification-summary:start -->
 
-`verify` invokes `format:check`, `lint`, `typecheck`, `check:audit`, `check:evidence`, `test`, `test:e2e:cli`, `test:e2e:readme`, `test:package`, `test:performance` in that order. `verify:full` runs that gate and the coverage suite.
+`verify` invokes `format:check`, `lint`, `typecheck`, `check:audit`, `check:evidence`, `test`, `test:e2e:cli`, `test:e2e:readme`, `test:e2e:coordination`, `test:package`, `test:performance` in that order. `verify:full` runs that gate and the coverage suite.
 
 <!-- loadout:verification-summary:end -->