Docs: replace the nonexistent hosted-Devin work-order command

[jeffhuber]

Part of stabilization epic #979 and roadmap #900.

Part of roadmap #900. Found while preparing the post-v1.4.0 hosted-Devin dogfood.

Problem

docs/devin-peer-support-qualification.md says hosted work orders run through a code-mower devin work-order command, but v1.4.0 exposes no such CLI. The maintained exact-bound path is the packaged DevinWorkOrders library seam documented in docs/devin-work-orders.md. A cold operator following the qualification guide therefore reaches a command that cannot run.

Outcome

All public guidance names an executable, supported path and makes the distinction between the generic remote-session CLI and exact PR-bound trusted work orders explicit.

Acceptance criteria

• Remove or correct every reference to the nonexistent command.
• Link the qualification guide directly to the current library embedding example.
• State that generic session dispatch provides provider-neutral lifecycle handling but does not independently verify issue, branch, PR, author, base, and exact head.
• If a CLI is desired, track it separately as a product feature rather than documenting it before it exists.
• Add a documentation contract test that every advertised top-level command parses under the packaged CLI.
• Keep credentials, identities, aliases, prompts, results, and private paths out of examples and test artifacts.

Delivery

One small independently reviewed documentation/test PR before the next documentation-bearing release.

Completion target and roadmap documentation

One planned OSS documentation/contract-test PR, planned Code Mower Claude builder with independent Code Mower Codex review, required before #915. Also synchronize the checked-in current-state/roadmap and qualification guidance with #900/#979: v1.4.0 released, bounded Devin builder qualification only, Graphify next, Board underway with #935 complete, supervised Slack planned. Preserve historical release/qualification records and mark unshipped runtime work as planned. #975 implements runtime role policy separately after these overlapping documentation edits; #955 owns effective-authority rendering.

[jeffhuber]

Work Order: #965 documentation lane scope

Code Mower Claude is the named builder for one independently reviewed PR closing #965. Use the maintained lane runner; Codex will provide the independent exact-head audit. Keep the PR draft until the orchestrator completes validation. Do not merge or release.

Own docs/devin-peer-support-qualification.md, docs/devin-work-orders.md, docs/current-state-and-roadmap.md, README.md current-scope wording, and a new tests/test_documented_commands.py. Do not change Graphify runtime/docs, review defaults, role-policy runtime, release reporting, docs/release-qualification.md, or historical release records. Avoid docs/sessions.md because its integration belongs to the Graphify lane.

Update to the approved #900/#979 plan and latest priority: fully close all seven v1.4.0 stabilization implementation children plus #974 evidence verification, ship those main-only fixes with Graphify v1.4.1 (#915), then Board v1.4.2. Published v1.4.0 artifacts remain immutable. Slack runtime work is deferred until that sequence is complete; supervised Slack remains planned for v1.5.0. Role-policy enforcement and other unshipped features must be described as planned, not implemented.

Preserve historical qualification tables and add the post-release campaign conclusion: 3/5 bounded hosted-builder deliveries; #865 and #935 completed separately by Codex #971/#973; the final #935 recovery was user-cancelled before delivery, not a completed implementation failure. Correct earlier invalid_request explanations: malformed or missing request keys caused the observed failures; valid same-session fix/cancel subsequently succeeded. Logical completion does not prove provider exit. Settled billing remains unknown.

Replace every advertised nonexistent hosted work-order command with the maintained packaged DevinWorkOrders library seam and direct embedding-example link. Explain generic session dispatch lifecycle versus exact PR-bound independent verification. Do not add a speculative CLI feature. Add a meaningful offline documentation contract test that parses the documented top-level commands with the packaged CLI, catches a fabricated command, and never launches provider or network work. Run focused tests plus existing documentation checks as appropriate.

Only sanitized repo/issue/PR/head, test, and audit metadata belongs in public delivery comments. Keep local paths, identities, credentials, raw output, private provider references, and prompts out of public artifacts.

[jeffhuber]

Mac lane runner (claude): this run ended without a validated delivery. The unit stays open for the next cycle.

• provider exit: 1
• supervision: completed
• observed transition: none
• classification: provider_exit_nonzero

[jeffhuber]

Mac lane runner (claude): this run ended without a validated delivery. The unit stays open for the next cycle.

• provider exit: 1
• supervision: completed
• observed transition: none
• classification: provider_exit_nonzero

[jeffhuber]

Code Mower Claude lane: delivered one draft PR closing #965.

• PR: Docs: replace the nonexistent hosted-Devin work-order command #981 (draft)
• Head SHA: 6c431ad6288477549daa317cc8917715f82eb9be
• Branch: claude/965-docs-work-order-command
• Labels: builder:claude, needs-codex-audit, documentation

Scope covered:

• Removed the advertised code-mower devin work-order command from
  docs/devin-peer-support-qualification.md and pointed the hosted rows at the
  packaged DevinWorkOrders library seam with a direct link to its embedding
  example anchor in docs/devin-work-orders.md.
• Stated that generic session dispatch provides provider-neutral lifecycle
  handling but does not independently verify issue, branch, PR, author, base,
  and exact head; only the exact PR-bound path does, and it fails closed.
• Recorded the post-release campaign conclusion (3/5 bounded hosted-builder
  deliveries; Builder delivery: honor target repository branch-name policy before PR creation #865 and Sessions: discover the current local brief and lease from a checkout #935 completed separately in Honor repository branch-name policy before builder PR creation #971/Sessions: discover and validate the current checkout brief #973; the final Sessions: discover the current local brief and lease from a checkout #935
  recovery was user-cancelled before delivery), corrected the earlier
  invalid_request explanation to malformed or missing request keys with valid
  same-session fix/cancel succeeding afterward, and stated that logical
  completion does not prove provider exit and settled billing remains unknown.
• Synchronized docs/current-state-and-roadmap.md and README current-scope
  wording with the approved sequence: stabilization children plus Cloud: verify aggregate freshness after accepted release metadata #974, then
  Graphify v1.4.1 (Release: qualify and publish optional Graphify context in v1.4.1 #915), then Board v1.4.2 with Sessions: discover the current local brief and lease from a checkout #935 complete, supervised Slack
  planned for v1.5.0; role-policy enforcement (Participants: enforce role-specific qualification for hosted builders #975) and effective-authority
  rendering (Adoption polish: report effective review authority and superseded Devin bridge drift #955) marked planned. Historical release records and qualification
  tables were preserved.
• No speculative CLI was added; a work-order CLI is called out as separate
  product work.
• Added tests/test_documented_commands.py: every top-level command advertised
  in current guidance must parse and dispatch under the packaged CLI, a
  fabricated command (including devin work-order) must fail to parse, and no
  current document may re-advertise the hosted work-order command. Handlers are
  replaced with recording stubs, so the test performs no provider or network
  work. Historical changelog and per-version release notes are excluded by
  design.

Tests run: none locally. Python execution was denied in this lane's sandbox
(/usr/bin/python3 was not permitted), so the focused suites and Ruff were not
executed here. CI on this head runs the new test, the existing documentation
checks, Ruff, and the full suite.

Remaining:

• Independent exact-head Codex audit.
• CI confirmation of the new contract test; if the documentation extractor needs
  tuning for a file this lane could not execute against, that is a fix round on
  the same branch.
• PR stays draft until the orchestrator completes validation. Not merged, not
  released.

[jeffhuber]

Implementation is ready for integration in draft PR #981 at 7cec22b0ffc54500d0165d4bcea2e86e3abd259d.

Code Mower Claude was the sole builder through the maintained local lane runner. The independent Code Mower Codex audit passed with no findings. Canonical validation passed 13 focused tests and 286 subtests, Ruff, public privacy scanning, and whitespace checks. Normal CI, including Python 3.12/3.13/3.14 package matrices and Linux/macOS graph containment, passed. The authoritative code-mower/gate status passed.

The builder process exited cleanly and its local coordination lease was released. The PR remains draft for roadmap integration; no merge or release has been performed. #965 closes when its PR is accepted and merged.

[jeffhuber]

Refreshed #981 through the existing Code Mower Claude lane by merging main at 529b70697908c4748106740fb241698783609fc9. Current head: 11e2b8d6c7ed5812d6ef301490450b82dfe0dc34.

The new independent Code Mower Codex audit passed with zero findings. Canonical tests passed 41 cases and 289 subtests, including documentation contracts/links/current-release hygiene and the merged review-default regression tests. Ruff, public privacy scanning, and whitespace checks passed. Normal CI and the authoritative code-mower/gate status passed on this head.

Claude remains the sole contributor to this PR's current diff. The maintenance process exited cleanly, and its local coordination lease is released. The PR remains draft for the orchestrator's integration decision; no merge or release was performed by this lane.