feat(brief): add pipeline wait rule to no-mistakes ship briefs

[TastyTom13]

Intent

Said 2026-09-29 evening: "we have been burning tokens at a rate faster then usual. I wonder why."

Context found by firstmate: the generated crew instructions say nothing about how to wait for the no-mistakes pipeline, so workers improvise. Pi workers ran sleep 180; no-mistakes axi status loops for hours, one model turn every three minutes per worker, which emptied the ChatGPT pool. A Claude worker put a change monitor on no-mistakes axi status, whose elapsed timers change every second, and received 60 events in 30 minutes, each a model turn.

What Changed

• bin/fm-dod-lib.sh: the generated no-mistakes crew brief now has a "Waiting on the pipeline" section. It tells workers to keep the drive call's --wait within their harness command limit and to wait with a single sleep 540 call between no-mistakes axi status checks. It forbids change monitors or watch tools on axi status output, because its timers change every second. It also tells workers to append a paused: awaiting pipeline <step> on <branch or PR> line to the status file, and to add a new line only when the step changes.
• tests/fm-brief.test.sh: no-mistakes briefs must contain the new section, the harness-bounded --wait rule and the paused declaration. Direct-PR and local-only briefs must not contain the section.

Risk Assessment

✅ Low: The change adds a short wait-rule section to the no-mistakes brief and a test that checks the generated brief; the earlier review fixes (verb from the shared variable, waits under the harness cap) are in place.

Testing

I ran the real brief generator in a throwaway home for all three modes and once with the paused verb overridden. The no-mistakes brief has the wait rule, the other two do not, and the override shows up in the brief. I also ran the whole brief test file and it passed. No UI is involved. I could not run a live worker to see if it obeys the text, because that needs a real model run.

• Live validation: ✅ go - 4 of 4 scenarios driven live against the product

Scenario	Result	Live	Evidence
Generate a no-mistakes brief: it has the wait section (bounded --wait, one sleep 540 between checks, no monitor on axi status, paused declaration)	✅ pass	live	brief-transcript.txt
Generate direct-PR and local-only briefs: they must not carry the wait section	✅ pass	live	brief-transcript.txt (grep count 0 for both)
Override FM_CLASSIFY_PAUSED_VERB=holding: the brief tells the worker to write holding: not paused:	✅ pass	live	brief-transcript.txt line 152 shows holding: declaration
All modes still generate cleanly, with no broken heredoc marker (EOF)	✅ pass	live	exit 0 for all modes, EOF count 0, fm-brief-test.txt

Evidence: Generated briefs across modes and verb override

Source: Generated briefs across modes and verb override

== no-mistakes: wait section ==
# Waiting on the pipeline
- Keep the drive call's `--wait` at or under your harness's command limit: the default `8m` fits a ten-minute cap, and a longer value gets the call killed.
- When a call returns with the run still going, wait with one call that fits the same limit, such as `sleep 540`, before the next `no-mistakes axi status`.
- Every check costs you a model turn, so never check more often than that.
- Never attach a change monitor or watch tool to `no-mistakes axi status` output: its timers change every second, so every tick wakes you for nothing.
- The pipeline never advances past a gate on its own, so a long wait loses nothing.
- Before each wait, append `paused: awaiting pipeline <step> on <branch or PR>` to the status file, and append a new line only when the step changes.

== direct-PR count ==
0
== local-only count ==
0
== override verb (holding) ==
152:- Before each wait, append `holding: awaiting pipeline <step> on <branch or PR>` to the status file, and append a new line only when the step changes.
== unterminated heredoc marker (EOF) count ==
0
== default verb ==
152:- Before each wait, append `paused: awaiting pipeline <step> on <branch or PR>` to the status file, and append a new line only when the step changes.

Evidence: fm-brief.test.sh output

Source: fm-brief.test.sh output

ok - fm-brief: scaffolds leave the worker role scope to the launch boundary and keep the secondmate contract
ok - fm-brief.sh: bash -n succeeds
/private/var/folders/08/gk0k1cz96snd794rnghc_98r0000gn/T/fm-brief.IpE8Nb/heredoc-in-substitution.sh:2
ok - fm-brief.sh: no heredoc is nested inside a command substitution (Bash 3.2 parse-safe)
ok - fm-brief.sh: --help renders the complete header
ok - fm-brief.sh: no-mistakes/direct-PR/local-only briefs generate cleanly
ok - fm-brief.sh: ship --mode is required and closed-set validated
ok - fm-brief.sh: the explicit ship mode wins over the registered posture
ok - fm-brief.sh: --yolo and scout/secondmate --mode are refused, never silently dropped
ok - fm-brief.sh: the base branch line renders only for an explicit base
ok - fm-brief.sh: faster paths use configured authority without stacked review
ok - fm-brief.sh: no-mistakes DOD keeps its apostrophe prose and bans --yes outright
ok - fm-brief.sh: no-mistakes ask-user findings use one event plus a verbatim snapshot
ok - fm-brief.sh: ship project-memory wording carries the AGENTS.md authoring bar
ok - fm-brief.sh: --herdr-lab emits the complete hard safety contract
ok - fm-brief.sh: --herdr-lab uses its quoted Firstmate-owned helper path
ok - fm-brief.sh: ship and scout scaffolds make omitted Herdr intent fail-visible
ok - fm-brief.sh: the documented {TASK} and {FIRSTMATE_SPEC} fills cannot corrupt the Herdr safety gate
ok - fm-brief.sh: Herdr lab contract covers scouts and rejects secondmate misuse
ok - fm-brief.sh: --no-projects scaffolds a project-less charter and guards misuse
ok - fm-brief.sh: marked requests avoid generic acknowledgements and preserve material reporting
ok - fm-brief.sh: relative directory inputs ignore CDPATH, render stable absolute charter paths, or fail loudly
ok - fm-brief.sh: custom pause verb renders in every scaffold
ok - fm-brief.sh: ship and scout scaffolds teach validation-round pauses
ok - fm-brief.sh: investigation and visual-review completions load the shared decision policy
ok - fm-brief: scout and secondmate code paths still scaffold well-formed briefs
ok - fm-brief.sh: scout Lavish hosting follows the bootstrap lavish-axi floor and names a durable board path
ok - fm-brief.sh: ship and scout carry the CONTEXT pointer, Toolkit, and Reporting rules
ok - fm-brief.sh: --env-file is opt-in, absolute, self-explaining, and refused on charters
ok - fm-brief.sh: the merge wait is declared exactly where a PR is raised
ok - fm-brief.sh: the Built-by PR-body line reads this task's own recorded harness/model/effort
ok - fm-brief.sh: fable-prompting Task subsections, charter Intent slot, Working discipline lines, cd caution, and ship verification step render for every scaffold
ok - fm-brief.sh: acceptance-numbering and red-then-green rules render only for PR-raising ship briefs
ok - fm-brief.sh: the screenshot pre-flight renders only for --ui tasks
ok - fm-brief.sh: --help documents the --ui screenshot pre-flight
ok - fm-brief.sh: ship and scout briefs name the unwanted and wanted turn ends plus the time sentence
ok - fm-brief.sh: --pasted-file marks the captain's paste under Captain's intent and is otherwise byte-identical
ok - fm-brief.sh: --design adds only the front-end design block
ok - fm-brief.sh: --visual is surface-bound, shared by ship/scout, and byte-neutral when absent
ok - fm-brief.sh: --help documents the visual-work flag and surface selector
ok - fm-brief.sh: --help documents --pasted-file and --design

Pipeline

Updates from git push no-mistakes

✅ **intent** - passed

✅ No issues found.

✅ **Rebase** - passed

✅ No issues found.

✅ **Review** - passed

✅ No issues found.

✅ **Test** - passed

✅ No issues found.

• Live validation: ✅ go - 4 of 4 scenarios driven live against the product

Scenario	Result	Live	Evidence
Generate a no-mistakes brief: it has the wait section (bounded --wait, one sleep 540 between checks, no monitor on axi status, paused declaration)	✅ pass	live	brief-transcript.txt
Generate direct-PR and local-only briefs: they must not carry the wait section	✅ pass	live	brief-transcript.txt (grep count 0 for both)
Override FM_CLASSIFY_PAUSED_VERB=holding: the brief tells the worker to write holding: not paused:	✅ pass	live	brief-transcript.txt line 152 shows holding: declaration
All modes still generate cleanly, with no broken heredoc marker (EOF)	✅ pass	live	exit 0 for all modes, EOF count 0, fm-brief-test.txt

• FM_HOME=<tmp> bin/fm-brief.sh b-<mode> some-proj --mode no-mistakes|direct-PR|local-only, then inspected the generated brief.md files
• FM_CLASSIFY_PAUSED_VERB=holding bin/fm-brief.sh ... --mode no-mistakes and checked the rendered declaration
• bash tests/fm-brief.test.sh (whole file, exit 0)

✅ **Document** - passed

✅ No issues found.

✅ **Lint** - passed

✅ No issues found.

✅ **Push** - passed

✅ No issues found.

Built by: claude/sonnet at medium