Fleet protocol
The contract. A lane's brief says what it works on; this says how it behaves.
Everything fleet-specific — repo names, lane names, paths, bases, hands-off
sets — lives in fleet.config.json (see references/configuration.md). This
file names no repository.
The protocol is a tool, not a reading assignment. Every ritual below has a
one-command form in ${CLAUDE_PLUGIN_ROOT}/scripts/fleet.mjs, and the
evidence rule that governs it is built into the command so it cannot be
skipped. Prefer the command to the hand-written version every time — the
hand-written versions are what cost the source run hours.
| Ritual |
Command |
| register + first heartbeat, one action |
fleet register <lane> <session> |
| heartbeat, validated on write, read back before success |
fleet hb <lane> <state> <task> [note] |
| what am I waiting on, has it cleared |
fleet blockers <lane> |
| every lane's state and age |
fleet census |
| a PR's check state, read properly |
fleet checks <owner/repo> <pr> |
| which posted verdicts have gone stale |
fleet verdicts <owner/repo> |
| who a merge unblocks and whose green it broke |
fleet unblocks <owner/repo> <pr> |
| CI load in jobs, not runs |
fleet queue-depth <owner/repo> |
| the router's digest |
fleet digest |
| a run directory from a config |
fleet init <run-id> · fleet doctor |
(fleet = node "${CLAUDE_PLUGIN_ROOT}/scripts/fleet.mjs"; --self-test
proves the pure logic on fixtures, including the double-run and mixed-type
check cases.)
Session mode. Each lane is its own session, not a subagent. A
foreground Agent call blocks the caller's turn: it cannot heartbeat and
cannot read messages until the call returns. Subagents are for bounded work
inside a lane's turn, never for a lane itself.
1. Registration and addressing
On start, resolve your own identity (ListAgents — its first line names
this session) and append one row to the run's registry.md:
| <UTC> | <lane-name> | <session-name> |. The session name is the
sidebar title and changes when a human renames it; the registry is
append-only, and the last row for a lane is its current address.
Resolve every address from the file at send time, never from memory.
Read the last registry row for that lane, then send. If a send fails to
resolve, run ListAgents once, match the row whose title fits the
role, append a fresh registry row, retry once. A role with no registry row
is not online — send to the dispatch router instead.
This is a rule about the reader, not the file. Two sessions independently
addressed a session that had been ruled hung hours earlier, in the same
hour, from memory — while the registry was correct the entire time.
One was caught by accident, the other by the first's confession, and no
process caught either. A correct, authoritative artifact that nobody
consults is the hardest instrument failure to fix, because there is
nothing to repair.
Never poll. Messages enqueue at the receiver and are processed on its
next turn, so a message wakes an idle session. A polling loop over
ListAgents burns turns and proves nothing.
Route through the dispatch router. All acks, reports, "PR ready"/"merged"
notices, blocked notices and escalations go to one comms role, which fans
out, collects acknowledgements, keeps the registry current, and sends the
orchestrator digests, not a stream. Nobody else messages tier 0 directly.
Two rules the router itself is bound by, both learned the hard way:
- Relay verbatim; a relay that sharpens is a relay that falsifies.
Paraphrase silently drops hedges. Adding a location, number or name the
source never supplied destroys the source's uncertainty while looking like
comprehension.
- Carry provenance across every hop ("per
<lane>, unverified"). A
router multiplies confidence without adding evidence: a claim that travels
lane → router → planner → back to its own origin arrives wearing two hops
of apparent independent corroboration it never had.
2. Heartbeats and standby
You own exactly one file: heartbeats/<your-lane-name>.md. Append one line:
<UTC> | <state> | <task> | <note>
state ∈ start working waiting blocked delivered standby. The
note is optional; the other three fields are not, and no field may contain an
unescaped newline. The line is not just a convention — validate it with
${CLAUDE_PLUGIN_ROOT}/scripts/validate-heartbeat.mjs. Three sessions
produced three different malformations of this format in one evening, and one
of them — a leading |, copied from a table — shifted every field so a
tombstone for a dead session made the monitor report it alive.
Append at start, at every milestone, before and after any command
expected to run long, and at least every heartbeat.intervalMinutes
(default 15) while active.
- A heartbeat older than
heartbeat.staleMinutes (default 25) whose state is
not standby is treated as hung — see the heartbeat-monitor skill for
what the monitor must read before acting on that.
- About to block on a foreground call? Heartbeat first:
waiting | <task> | foreground worker, expect silence. The monitor reads
it as intended rather than as a hang.
waiting and standby are different states, and only one has a
counterparty. waiting | <specific next action> | <why> is reserved for
something genuinely pending on someone else, and it names what — so the
monitor, and you, can check whether that thing is still pending.
standby | <next action when work arrives> | is genuinely idle. Collapsing
them makes a blocked lane and an idle lane identical in a census; a lane
once sat blocked for two hours after its blocker had cleared.
- Whatever you are waiting on, measure it yourself each heartbeat:
fleet blockers <your-lane>. It reads the refs in your own waiting
heartbeats and asks the forge whether each has cleared — exit 1 means
move. The merger telling you is the other half of that rule and it is
unauditable — it leaves no trace, so nobody can verify it was sent,
including the merger. Your own measurement leaves one, and it caught 3 of 3
idle-on-done-work cases that no monitor saw.
- Idle with nothing queued: write
standby and end your turn. A message
wakes you. Standby is not death.
- Your predecessor (a replaced session of the same lane name) left its last
state in that file. Resume from it; do not restart its work.
Appending: these files are written by many sessions at once. Use fleet hb.
It prints on every branch and reads the line back before reporting
success, because a shared helper that no-ops quietly is worse than twenty
lanes appending by hand — it fails them all in the same silent way, and a
silent heartbeat skip reads as HUNG and gets a producing lane replaced.
"Not written", "denied" and "skipped" leave identical bytes; only the
read-back tells them apart. If you must append by hand: on Windows,
[IO.File]::Open($p,'Append','Write','ReadWrite') or Add-Content; POSIX
>> and tail against a file another session holds open can block until the
tool timeout — and read your own line back afterwards. See
docs/platform-notes.md.
3. ASSERTED vs DELIVERED
Every statement a session makes is ASSERTED until a link promotes it to
DELIVERED.
A promoting link is a PR URL, an issue number, a merge SHA, a re-run of the
exact command by a different agent, or a check state read at a named head.
Not promoting: "done", "implemented", "should be fine", a worker's report of
its own success, a plan, or an intention.
- A worker's "done" is a claim until a verifier re-runs the stated command.
- A count produced by a pattern match over a file is not a measurement of the
thing counted.
- The claim state is recorded in the run's logs — the queue, the ack log and
the review log each carry it — so a later reader can tell which claims were
ever checked.
This is the single highest-value rule in the protocol. Everything in the
evidence-rules skill is a corollary of it.
4. Git discipline
Full text with the incident behind each rule: references/git-discipline.md.
The short form, all of it load-bearing:
- One repo per lane, exclusive scope, from
fleet.config.json. Never
edit another lane's repo. Honour the configured hands-off repo list.
- Never switch a shared checkout's HEAD, never commit in it, never stash
it. Work in a per-task worktree. If a predecessor's worktree already holds
your branch, continue there — one writer per branch.
- Push and open the PR in ONE action. A pushed branch with no PR is a
parking spot that hides work. Draft is the parking spot; a bare branch is
not.
- Prove on the committed tree, not the worktree:
git diff HEAD --stat
must be empty when the proof runs.
- Conventional commits, with a
Co-Authored-By: trailer naming the
model. No authorship headers inside source files.
- Never
--force, reset --hard, clean, or worktree remove --force.
Never delete a branch holding unique content. --force-with-lease is
permitted only on a lane-owned branch and only in the form
--force-with-lease=<ref>:<sha> with the SHA written out — a bare lease
reads a remote-tracking ref that a background fetch may already have
moved, which makes it a guard-shaped no-op.
- Fix root causes. Never weaken, skip, quarantine or retry-loop a test to
get green. A green test that enshrines the defect is a defect.
- No machine-absolute paths and no time estimates in committed files.
Sequence by dependency.
- Verify the issue's symptom before building. Open state is not evidence
the defect exists — and presence of the file, script or component is not
presence of the defect. Absent → close the issue with the evidence.
- Do not act on a PR you merely discovered by scanning. In a fleet
sharing one credential,
author cannot discriminate between lanes; it
answers "is this ours", and the branch answers "which lane". Read
authorship from the branch, never from the PR.
5. Review and merge-readiness discipline
Review routing. Mechanical changes — docs, templates, link fixes, and
anything whose correctness is decided by a gate the author already ran — go
to a peer lane, preferably one that has run the same gate. Management
review is reserved for product behaviour, security, permissions, gates
themselves, and anything on a release path. The peer reviewer is bound by
every rule that binds a management reviewer. Never route a lane to review its
own branch: self-approval wearing a peer's clothes.
Every PR body states four things — this is the shape that makes reviewer
attention go further, because the claims arrive already scoped:
|
|
| WHAT was verified |
the claim |
| THE EXACT COMMAND or gate |
so the claim is re-runnable |
| THE FALSIFIER |
the input that would have turned it red |
| THE SET or COUNT |
what the claim ranges over |
Plus: what was intentionally not run, and the rejected "fixes" that would
have hidden the problem rather than solved it.
Before merging (references/merge-discipline.md for the mechanics):
- Re-read the checks as the literal last action. Run the unfiltered
pending-check query immediately before the merge command. Never reuse a
result from earlier in the same turn. Stop on any contradiction between
two reads. Incident: a merge went through with two checks still queued
because a filtered query returned empty while an unfiltered list in the
same message showed both pending.
- A SHA handed to another session is read fresh from the remote
immediately before handing it on, never from the last local push. The
verdict, the merge and the last look are all scoped to that SHA; a stale
one silently scopes all three to a tree nobody read.
- A squash merge makes a new SHA, so the merge commit legitimately
carries no checks. The verdict lives on the PR head. "The merge commit has
no checks" is never itself evidence of a gap.
- A structural readiness verdict is not CI. Whatever your merge-gate
tool reports about mergeability, absorption or worktree cleanliness, it
read no check conclusions unless it says it did.
6. Escalation
Some decisions are not a session's to make: money, production, tenants, the
trust chain, repository or organisation settings, abandoning work. Configure
the list in fleet.config.json under founderClass.
- An escalation parks a decision, never a lane. Append one line to
escalations.md — <UTC> | <lane> | <decision needed> | MEANWHILE: <what proceeds> — notify the dispatch router, and continue with the next task.
- Reaching the human means a ready-to-merge PR with proofs, not a
question that parks the work.
- Never ask the human a blocking question mid-run.
6a. Classify before you escalate
founderClass names topics. It does not decide who acts. Classify each
blocker from facts, using the table in references/authority-and-receipts.md
(the checker is lib/fleet/classify-authority.mjs):
| Class |
When |
Goes to |
| IRREVERSIBLE/FOUNDER |
Merge to default branch, publish, tenant-wide or production grant, billing, deleting customer data, production cutover |
coordinator → parent → founder |
| MISSING-CAPABILITY |
Reversible, but no automated path or permission exists |
coordinator, naming the exact permission or capability |
| AGENT-AUTHORIZED |
Reversible and inside an existing grant or the brief |
proceed, no sign-off |
| COORDINATOR-DECISION |
Reversible and capable, but not delegated |
coordinator extends or denies the brief |
The founder is reached only through the parent, and only for the first row.
Do not list a reversible step the brief delegates as "the human's step".
6b. Receipts
Every escalation, completion and failure is a schema-checked receipt, sent to
the coordinator's thread (references/authority-and-receipts.md). A lane is
done only when it has posted a completion or failure receipt. A cancelled
run still owes one. The coordinator lists terminal runs that have no receipt
and reconciles every completion claim against T3 run state.
7. Behaviour
- Minimum verbosity in every message and report: facts, numbers,
path:line.
- Delegate. Spend your own turns on judgement, routing and verification.
Bounded edits go to a mid-tier worker; verification to a cheap verifier.
Every dispatch names its model — see the
fleet-roles skill.
- Spawn workers in the background and heartbeat while they run.
- Loop until the planner says the queue is empty, then write
reports/<lane>.md (PRs opened and merged, issues closed, blocked items,
exact proofs) and finish with heartbeat standby.
Run directory
One directory per run, machine-local, never committed. Layout, ownership and
the append rule: references/run-directory.md.
References
| File |
Contents |
references/configuration.md |
Every configurable name, and the config file shape |
references/git-discipline.md |
The full git rules with the incident behind each |
references/merge-discipline.md |
The merge gate, step by step |
references/run-directory.md |
Log files, who owns which, the append rule |
references/authority-and-receipts.md |
Blocker decision table, escalation receipt schema, done rule, coordinator reconciliation |
1---2name: fleet-protocol3description: The contract every session in a multi-session Claude Code fleet reads first. Use when starting, joining, or running a fleet of parallel sessions on one codebase - heartbeats and standby, addressing other sessions through a dispatch router, ASSERTED vs DELIVERED claim states, git discipline (one repo per lane, push-and-open-PR as one action, never rewrite shared history), and merge-readiness discipline (re-read the checks as the literal last action before a merge). Triggers on "start a fleet", "orchestration protocol", "lane", "heartbeat", "dispatch", "PR ready", "merge gate", "who owns this repo", or any question about how parallel sessions coordinate.4metadata:5 source: https://skillmd.com/skills/lobbi-docs/fleet-protocol6---78# Fleet protocol910The contract. A lane's brief says what it works on; this says how it behaves.11Everything fleet-specific — repo names, lane names, paths, bases, hands-off12sets — lives in `fleet.config.json` (see `references/configuration.md`). This13file names no repository.1415**The protocol is a tool, not a reading assignment.** Every ritual below has a16one-command form in `${CLAUDE_PLUGIN_ROOT}/scripts/fleet.mjs`, and the17evidence rule that governs it is built into the command so it cannot be18skipped. Prefer the command to the hand-written version every time — the19hand-written versions are what cost the source run hours.2021| Ritual | Command |22|---|---|23| register + first heartbeat, one action | `fleet register <lane> <session>` |24| heartbeat, validated on write, **read back before success** | `fleet hb <lane> <state> <task> [note]` |25| what am I waiting on, has it cleared | `fleet blockers <lane>` |26| every lane's state and age | `fleet census` |27| a PR's check state, read properly | `fleet checks <owner/repo> <pr>` |28| which posted verdicts have gone stale | `fleet verdicts <owner/repo>` |29| who a merge unblocks and whose green it broke | `fleet unblocks <owner/repo> <pr>` |30| CI load in jobs, not runs | `fleet queue-depth <owner/repo>` |31| the router's digest | `fleet digest` |32| a run directory from a config | `fleet init <run-id>` · `fleet doctor` |3334(`fleet` = `node "${CLAUDE_PLUGIN_ROOT}/scripts/fleet.mjs"`; `--self-test`35proves the pure logic on fixtures, including the double-run and mixed-type36check cases.)3738**Session mode.** Each lane is its **own session**, not a subagent. A39foreground `Agent` call blocks the caller's turn: it cannot heartbeat and40cannot read messages until the call returns. Subagents are for bounded work41*inside* a lane's turn, never for a lane itself.4243## 1. Registration and addressing44451. On start, resolve your own identity (`ListAgents` — its first line names46 this session) and append one row to the run's `registry.md`:47 `| <UTC> | <lane-name> | <session-name> |`. The session name is the48 sidebar title and changes when a human renames it; the registry is49 append-only, and the **last** row for a lane is its current address.502. **Resolve every address from the file at send time, never from memory.**51 Read the last registry row for that lane, then send. If a send fails to52 resolve, run `ListAgents` **once**, match the row whose title fits the53 role, append a fresh registry row, retry once. A role with no registry row54 is not online — send to the dispatch router instead.5556 This is a rule about the reader, not the file. Two sessions independently57 addressed a session that had been ruled hung hours earlier, in the same58 hour, from memory — while the registry was **correct the entire time**.59 One was caught by accident, the other by the first's confession, and no60 process caught either. A correct, authoritative artifact that nobody61 consults is the hardest instrument failure to fix, because there is62 nothing to repair.633. **Never poll.** Messages enqueue at the receiver and are processed on its64 next turn, so a message *wakes* an idle session. A polling loop over65 `ListAgents` burns turns and proves nothing.6667**Route through the dispatch router.** All acks, reports, "PR ready"/"merged"68notices, blocked notices and escalations go to one comms role, which fans69out, collects acknowledgements, keeps the registry current, and sends the70orchestrator **digests, not a stream**. Nobody else messages tier 0 directly.7172Two rules the router itself is bound by, both learned the hard way:7374- **Relay verbatim; a relay that sharpens is a relay that falsifies.**75 Paraphrase silently drops hedges. Adding a location, number or name the76 source never supplied destroys the source's uncertainty while looking like77 comprehension.78- **Carry provenance across every hop** ("per `<lane>`, unverified"). A79 router multiplies confidence without adding evidence: a claim that travels80 lane → router → planner → back to its own origin arrives wearing two hops81 of apparent independent corroboration it never had.8283## 2. Heartbeats and standby8485You own exactly one file: `heartbeats/<your-lane-name>.md`. Append one line:8687```88<UTC> | <state> | <task> | <note>89```9091`state` ∈ `start` `working` `waiting` `blocked` `delivered` `standby`. The92note is optional; the other three fields are not, and no field may contain an93unescaped newline. **The line is not just a convention — validate it** with94`${CLAUDE_PLUGIN_ROOT}/scripts/validate-heartbeat.mjs`. Three sessions95produced three different malformations of this format in one evening, and one96of them — a leading `|`, copied from a table — shifted every field so a97tombstone for a dead session made the monitor report it **alive**.9899Append at start, at every milestone, **before and after** any command100expected to run long, and at least every `heartbeat.intervalMinutes`101(default 15) while active.102103- A heartbeat older than `heartbeat.staleMinutes` (default 25) whose state is104 not `standby` is treated as hung — see the `heartbeat-monitor` skill for105 what the monitor must read before acting on that.106- About to block on a foreground call? Heartbeat **first**:107 `waiting | <task> | foreground worker, expect silence`. The monitor reads108 it as intended rather than as a hang.109- **`waiting` and `standby` are different states, and only one has a110 counterparty.** `waiting | <specific next action> | <why>` is reserved for111 something genuinely pending on someone else, and it names what — so the112 monitor, and you, can check whether that thing is still pending.113 `standby | <next action when work arrives> |` is genuinely idle. Collapsing114 them makes a blocked lane and an idle lane identical in a census; a lane115 once sat blocked for **two hours** after its blocker had cleared.116- **Whatever you are waiting on, measure it yourself each heartbeat:**117 `fleet blockers <your-lane>`. It reads the refs in your own `waiting`118 heartbeats and asks the forge whether each has cleared — exit 1 means119 move. The merger telling you is the other half of that rule and it is120 unauditable — it leaves no trace, so nobody can verify it was sent,121 including the merger. Your own measurement leaves one, and it caught 3 of 3122 idle-on-done-work cases that no monitor saw.123- Idle with nothing queued: write `standby` and **end your turn**. A message124 wakes you. Standby is not death.125- Your predecessor (a replaced session of the same lane name) left its last126 state in that file. Resume from it; do not restart its work.127128Appending: these files are written by many sessions at once. Use `fleet hb`.129It prints on **every** branch and **reads the line back** before reporting130success, because a shared helper that no-ops quietly is worse than twenty131lanes appending by hand — it fails them all in the same silent way, and a132silent heartbeat skip reads as HUNG and gets a producing lane replaced.133"Not written", "denied" and "skipped" leave identical bytes; only the134read-back tells them apart. If you must append by hand: on Windows,135`[IO.File]::Open($p,'Append','Write','ReadWrite')` or `Add-Content`; POSIX136`>>` and `tail` against a file another session holds open can block until the137tool timeout — and read your own line back afterwards. See138`docs/platform-notes.md`.139140## 3. ASSERTED vs DELIVERED141142> Every statement a session makes is **ASSERTED** until a link promotes it to143> **DELIVERED**.144145A promoting link is a PR URL, an issue number, a merge SHA, a re-run of the146exact command by a different agent, or a check state read at a named head.147Not promoting: "done", "implemented", "should be fine", a worker's report of148its own success, a plan, or an intention.149150- A worker's "done" is a claim until a verifier re-runs the stated command.151- A count produced by a pattern match over a file is not a measurement of the152 thing counted.153- The claim state is recorded in the run's logs — the queue, the ack log and154 the review log each carry it — so a later reader can tell which claims were155 ever checked.156157This is the single highest-value rule in the protocol. Everything in the158`evidence-rules` skill is a corollary of it.159160## 4. Git discipline161162Full text with the incident behind each rule: `references/git-discipline.md`.163The short form, all of it load-bearing:1641651. **One repo per lane**, exclusive scope, from `fleet.config.json`. Never166 edit another lane's repo. Honour the configured hands-off repo list.1672. **Never switch a shared checkout's HEAD**, never commit in it, never stash168 it. Work in a per-task worktree. If a predecessor's worktree already holds169 your branch, continue **there** — one writer per branch.1703. **Push and open the PR in ONE action.** A pushed branch with no PR is a171 parking spot that hides work. Draft is the parking spot; a bare branch is172 not.1734. **Prove on the committed tree**, not the worktree: `git diff HEAD --stat`174 must be empty when the proof runs.1755. **Conventional commits**, with a `Co-Authored-By:` trailer naming the176 model. No authorship headers inside source files.1776. **Never** `--force`, `reset --hard`, `clean`, or `worktree remove --force`.178 Never delete a branch holding unique content. `--force-with-lease` is179 permitted only on a lane-owned branch and only in the form180 `--force-with-lease=<ref>:<sha>` with the SHA written out — a bare lease181 reads a remote-tracking ref that a background fetch may already have182 moved, which makes it a guard-shaped no-op.1837. **Fix root causes.** Never weaken, skip, quarantine or retry-loop a test to184 get green. A green test that enshrines the defect is a defect.1858. **No machine-absolute paths and no time estimates** in committed files.186 Sequence by dependency.1879. **Verify the issue's symptom before building.** Open state is not evidence188 the defect exists — and presence of the file, script or component is not189 presence of the defect. Absent → close the issue with the evidence.19010. **Do not act on a PR you merely discovered by scanning.** In a fleet191 sharing one credential, `author` cannot discriminate between lanes; it192 answers "is this ours", and the branch answers "which lane". Read193 authorship from the **branch**, never from the PR.194195## 5. Review and merge-readiness discipline196197**Review routing.** Mechanical changes — docs, templates, link fixes, and198anything whose correctness is decided by a gate the author already ran — go199to a **peer lane**, preferably one that has run the same gate. Management200review is reserved for product behaviour, security, permissions, gates201themselves, and anything on a release path. The peer reviewer is bound by202every rule that binds a management reviewer. Never route a lane to review its203own branch: self-approval wearing a peer's clothes.204205**Every PR body states four things** — this is the shape that makes reviewer206attention go further, because the claims arrive already scoped:207208| | |209|---|---|210| WHAT was verified | the claim |211| THE EXACT COMMAND or gate | so the claim is re-runnable |212| THE FALSIFIER | the input that would have turned it red |213| THE SET or COUNT | what the claim ranges over |214215Plus: what was intentionally **not** run, and the rejected "fixes" that would216have hidden the problem rather than solved it.217218**Before merging** (`references/merge-discipline.md` for the mechanics):2192201. **Re-read the checks as the literal last action.** Run the *unfiltered*221 pending-check query immediately before the merge command. Never reuse a222 result from earlier in the same turn. Stop on any contradiction between223 two reads. Incident: a merge went through with two checks still queued224 because a filtered query returned empty while an unfiltered list in the225 same message showed both pending.2262. **A SHA handed to another session is read fresh** from the remote227 immediately before handing it on, never from the last local push. The228 verdict, the merge and the last look are all scoped to that SHA; a stale229 one silently scopes all three to a tree nobody read.2303. **A squash merge makes a new SHA**, so the merge commit legitimately231 carries no checks. The verdict lives on the PR head. "The merge commit has232 no checks" is never itself evidence of a gap.2334. **A structural readiness verdict is not CI.** Whatever your merge-gate234 tool reports about mergeability, absorption or worktree cleanliness, it235 read no check conclusions unless it says it did.236237## 6. Escalation238239Some decisions are not a session's to make: money, production, tenants, the240trust chain, repository or organisation settings, abandoning work. Configure241the list in `fleet.config.json` under `founderClass`.242243- **An escalation parks a decision, never a lane.** Append one line to244 `escalations.md` — `<UTC> | <lane> | <decision needed> | MEANWHILE: <what245 proceeds>` — notify the dispatch router, and continue with the next task.246- Reaching the human means a **ready-to-merge PR with proofs**, not a247 question that parks the work.248- Never ask the human a blocking question mid-run.249250### 6a. Classify before you escalate251252`founderClass` names *topics*. It does not decide who acts. Classify each253blocker from facts, using the table in `references/authority-and-receipts.md`254(the checker is `lib/fleet/classify-authority.mjs`):255256| Class | When | Goes to |257|---|---|---|258| IRREVERSIBLE/FOUNDER | Merge to default branch, publish, tenant-wide or production grant, billing, deleting customer data, production cutover | coordinator → parent → founder |259| MISSING-CAPABILITY | Reversible, but no automated path or permission exists | coordinator, naming the **exact** permission or capability |260| AGENT-AUTHORIZED | Reversible and inside an existing grant or the brief | **proceed**, no sign-off |261| COORDINATOR-DECISION | Reversible and capable, but not delegated | coordinator extends or denies the brief |262263The founder is reached only through the parent, and only for the first row.264Do not list a reversible step the brief delegates as "the human's step".265266### 6b. Receipts267268Every escalation, completion and failure is a schema-checked receipt, sent to269the coordinator's thread (`references/authority-and-receipts.md`). **A lane is270done only when it has posted a completion or failure receipt.** A cancelled271run still owes one. The coordinator lists terminal runs that have no receipt272and reconciles every completion claim against T3 run state.273274## 7. Behaviour275276- **Minimum verbosity** in every message and report: facts, numbers,277 `path:line`.278- **Delegate.** Spend your own turns on judgement, routing and verification.279 Bounded edits go to a mid-tier worker; verification to a cheap verifier.280 Every dispatch names its model — see the `fleet-roles` skill.281- **Spawn workers in the background** and heartbeat while they run.282- **Loop** until the planner says the queue is empty, then write283 `reports/<lane>.md` (PRs opened and merged, issues closed, blocked items,284 exact proofs) and finish with heartbeat `standby`.285286## Run directory287288One directory per run, machine-local, never committed. Layout, ownership and289the append rule: `references/run-directory.md`.290291## References292293| File | Contents |294|---|---|295| `references/configuration.md` | Every configurable name, and the config file shape |296| `references/git-discipline.md` | The full git rules with the incident behind each |297| `references/merge-discipline.md` | The merge gate, step by step |298| `references/run-directory.md` | Log files, who owns which, the append rule |299| `references/authority-and-receipts.md` | Blocker decision table, escalation receipt schema, done rule, coordinator reconciliation |