wave skill: the agents' standing rules move into the repo

.claude/skills/wave/agent-rules.md replaces /tmp/salvage/standing-rules.md.
SKILL.md and AGENTS.md point to it instead of restating it.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VdfEGXA7aJ6ftMsCc9PJze
This commit is contained in:
Edward Langley
2026-09-22 21:00:27 -07:00
parent 40b8913397
commit 7ce778fca2
3 changed files with 203 additions and 23 deletions
+25 -23
View File
@@ -33,32 +33,34 @@ it must cross. Two agents adding a workspace member will still collide in
## What every brief must carry
**Point the agent at [`agent-rules.md`](agent-rules.md) and tell it to read it in
full first.** That file is the agent's standing rules: base and push mechanics,
the gate list, the claim, the vacuity shapes, evidence, the files it must not
write, GitHub conventions, the hand-back and disk cleanup. Do not restate them
in the brief. A restated rule drifts from its source, and the brief is the copy
nobody corrects. The brief adds only what varies per wave:
1. **The expected base commit, by name.** Worktrees do not reliably branch from
HEAD — one was created three merges behind and only noticed because it was
told to check.
2. **The baseline**, so "N passing" is checkable rather than asserted: Rust test
count, harness counts, node counts.
3. **Its ground, and the ground that is not.**
4. **Never write** `.beads/**`, `prolog-sessions/kb.pl`, `context/repo-map.md`.
Agents read them freely and report proposed text; the orchestrator applies it.
5. **The commit trailers and PR footer**, spelled out.
6. **Do not push to `main`, do not merge.** The orchestrator merges.
7. **What to report**: branch and head SHA, PR number, what changed per ticket,
*what could not be verified and why*, proposed repo-map/kb text, and **what it
learned that the ticket did not predict** — consistently the most valuable
part, and omitted unless asked for explicitly.
HEAD. One was created three merges behind and was only noticed because it
was told to check.
2. **A reference baseline**, so "N passing" is checkable: Rust test count,
harness counts, node counts, and the machine. The agent still measures its
own.
3. **Its ground, and the ground that is not.** Name any orchestrator-only file
it is exceptionally allowed to edit.
4. **Its `BEADS_ACTOR` and its tickets.**
5. **The commit trailers and PR footer**, spelled out. They carry this
session's URL, which is why `agent-rules.md` cannot hold them.
## Method to require, because this repo's defect is a test that passes
When a hand-back teaches a new rule, add it to `agent-rules.md` in the same
wave, not to the next brief.
- Failing test first, then the fix.
- Every guard proved by **injecting the fault it names** and watching it fail
**by name**. An injection that does not compile reads as a clean run, so
require a positive signal the build ran.
- `cargo test --workspace` is **fail-fast**; `--no-fail-fast` for any claim about
blast radius.
- A textually clean merge can **delete tests** with no conflict marker, and
`cargo check` will not notice. After any merge that moves a test module, diff
test **names** — a count is masked by the other side's additions.
## Method to require
`agent-rules.md`, "Discipline per ticket" and "Gates": failing test first, a
rival injected and watched failing by name, `--no-fail-fast`, and test *names*
diffed after any merge. Hold the hand-back to it. This repo's characteristic
defect is a test that passes.
## Shepherding
+170
View File
@@ -0,0 +1,170 @@
# Standing rules for a wave agent
Every work-unit agent reads this **in full before its first command**. A brief
points here and adds only what is specific to its wave: the expected base
commit, the agent's ground, its `BEADS_ACTOR`, and its tickets. Every rule below
was bought by a failure.
This file holds each rule **once**. Where `AGENTS.md` ("The Wave Workflow")
already carries the history and the argument, the rule here is one line and
points there. Change a rule here, not in a brief, and not in a second copy.
## Base, branch, push
1. **Branch from `gh/main`, never `main` or `origin/main`.** `git fetch gh &&
git checkout -b <branch> gh/main`. A worktree handed to you may already sit
on unpushed commits. Why: `AGENTS.md`, "Shipping a wave" (the two remotes
both spelled `main`, and the worktree-base vector).
2. **Before opening a PR, check `git diff --numstat gh/main...<branch>`.** Do
not use `gh pr view --json files`: that is the union of your commits' files,
not the effective diff.
3. **Push with an explicit refspec:** `git push gh HEAD:refs/heads/<branch>`. A
branch cut from `gh/main` tracks `gh/main`, so a bare push targets `main` and
is refused.
4. **Before pushing a fix to an existing PR, check `gh pr view <n> --json
state`.** GitHub deletes a merged PR's branch. A push to it then prints
`* [new branch]` and lands nowhere. Why: `AGENTS.md`, "Shipping a wave".
## Commit early: a worktree is not durable storage
Commit a WIP as soon as your first test run is green, and again after every
meaningful step. Push your branch (rule 3) once it is worth keeping. An agent
finished a whole ticket with every gate green and lost all of it, uncommitted,
when its worktree was deleted in a disk cleanup. A local commit survives a
deleted worktree. Tidying history before the PR is fine; losing it is not.
## Gates
```
direnv exec . cargo test --workspace --no-fail-fast
direnv exec . cargo clippy --workspace --all-targets -- -D warnings
direnv exec . ./scripts/fmt-workspaces.sh # NOT cargo fmt --all; exit 2 = the gate did not run
direnv exec . python3 scripts/check-comment-runs.py --check-evidence
direnv exec . ./scripts/checkers.sh
direnv exec . ./scripts/node-test.sh 'web/tests/*.test.js'
```
- **Run everything under `direnv exec .`, and never pipe a gate.** Why, and
why not `cargo fmt --all`: `AGENTS.md`, brief item 3.
- **`--no-fail-fast`**, because plain `cargo test --workspace` stops at the
first failing binary. Use it for any claim about blast radius.
- If you touch a browser-bundle crate (core, formula, io, protocol, command,
wasm-client, worker-server), also run
`cargo check --target wasm32-unknown-unknown -p <crate>` **for each crate**.
CI checks each one separately, so feature unification cannot hide a leak.
- If you touch `web/` or `harness/`, run the Playwright suite in `harness/`.
- **Measure your own baseline** before changing anything, and report it with
the machine it was measured on. A figure from the brief is a reference, not
your baseline.
- **Merge `gh/main` and re-run every gate before calling a PR green.** Then
check that no test *disappeared*. Diff test **names**, not counts: a
textually clean merge can delete a test module with no conflict marker, and
the other side's new tests mask the loss in a count.
- **The shell guard refusing a compound command means NEITHER part ran.** Split
the command. After a refusal, re-check the state it was meant to produce.
- **Re-run the full suite after reverting an injected fault.** A revert applied
to the wrong occurrence once shipped two faults past every other gate.
## Discipline per ticket
- **Claim first:** `export BEADS_ACTOR=<your actor>` in every shell, then
`bd update <id> --claim`. Exit 13 means another actor holds it. Pick another;
do not retry. `bd` in a worktree needs `.envrc` allowed: `AGENTS.md`, the
paragraph after the brief item list.
- **Reproduce at HEAD before fixing.** A dozen tickets were stale-open in one
campaign, one fixed two days before it was filed. If it does not reproduce,
find the fixing commit, confirm with `git merge-base --is-ancestor <sha> HEAD`,
report the SHA, and close nothing.
- **Write a failing test first, then inject a rival and watch it fail.** Run
it; do not reason about it. Report *where* it failed, by case name. An
injection that does not compile reads as a clean run, so get a positive
signal that the build ran.
- **Check your test against every vacuity shape seen here:**
- an assertion the default state already satisfies;
- `f(x) == f(x)`: calling a pure `Cmd` twice against one immutable context;
- a disjunctive assertion whose other branch is the bug's own symptom;
- a rival that fails the *setup*, which proves the neighbouring question
(e.g. setup and subject share an `EffectKind`);
- a fixture that is right by accident (expected value 1, group at row 0,
bystander already on the relayed value);
- a fixture or stub that lacks the structure the hazard needs, or fails in a
wider shape than the real fault;
- hand-picked fixture cases where the set could be **enumerated**: four
vacuities in `fmt-workspaces-test.sh` were each a case nobody thought of
(improvise-90ea);
- a rendered-output test at a width where the thing asserted is never drawn;
- a fold test that batches differently from the real producer (a
`Binding::Sequence` is ONE effect vec);
- **the cursor at index 0**: "the item under the cursor" and "the first item"
are indistinguishable there, and that was four of five findings in one
review. Seat the cursor at a nonzero index of a list with at least two
elements;
- a determinism check against ONE context instance: every `HashMap` iterates
the same way twice. Build two equal contexts independently
(improvise-hzjo);
- two derivations of one fact where the fixture makes them agree (drawn vs
committed, edit text vs value, `1200` vs `1.23456789`). Use a value where
they disagree.
- **Record each bug in `context/test-evidence/<area>.md`**, keyed by test name
(module-qualify the name when the leaf is not unique). It is append-at-EOF, so
**expect a conflict**. Keep both sides and verify line by line against both
parents: `--check-evidence` cannot see a dropped entry.
- **Comment runs are at most two lines.** Reasoning goes in `context/`, not in
comments (`context/design-principles.md` §6).
- **If the acceptance criteria look wrong, argue it** rather than following
them silently. Three agents inverted theirs in one campaign, and all three
were right.
## Files you must not write
`.beads/**` (never commit `issues.jsonl`, never `git checkout` it),
`prolog-sessions/kb.pl`, `context/repo-map.md`: `AGENTS.md`, "Who owns what".
Also `AGENTS.md`, unless your brief grants you a named section. Report proposed
text for any of these in your hand-back.
## GitHub
- `gh` is shell-aliased to a different tool. Use
`nix shell nixpkgs#gh --command gh …`, and pass bodies with `-F body=@file`.
- **Everything posted goes out under the repository owner's account.** Open
every PR body and comment with:
`_Posted by Claude (<model>), an agent in Edward Langley's session._`
- Commit trailers and the PR footer are given by your brief, verbatim. They
carry a session URL, so they cannot live here.
- Read reviews with `./scripts/pr-reviews <n>`, and do not pipe it. How to tell
a ChatGPT Codex review from ours, and why Copilot is not coming: `AGENTS.md`,
"Shipping a wave". Answer each Codex review on its thread.
- **Do not merge, and do not close beads.** Avoid stacked PRs. If you must
stack, never let the base branch be deleted.
## Hand-back, per ticket
Report:
- the claim result;
- your measured baseline;
- whether the ticket reproduced (the SHA if it did not);
- the rival and where it failed;
- what you could not verify, and why;
- proposed repo-map and `kb.pl` text;
- **what the ticket did not predict**, consistently the most valuable part.
## Disk
Each worktree's `target/` grows to 3–6G, and a full disk (ENOSPC) once stopped
every agent. **When you hand back, delete your own build output, and only
yours.**
Run `pwd` and `git rev-parse --show-toplevel` first, and read the answer. Then
delete by **literal absolute path**, typed out:
`rm -rf /Users/.../.claude/worktrees/agent-<you>/target`.
**Never interpolate a computed path into `rm -rf`.** One agent's worktree had
been removed, so `$(git rev-parse --show-toplevel)` printed nothing and the
command became `rm -rf "/target"`. A guarded one-liner was tried, but the
worktree shell guard refuses it as too complex, which pushed agents back to an
unguarded form. A literal path is the version that is both allowed and safe.
Keep your branch and any uncommitted work. Delete only `target/` (and
`harness/opfs-contract/target` if you built it). If you see ENOSPC, stop and
report rather than deleting anything outside your worktree.
+8
View File
@@ -100,6 +100,14 @@ Two agents each adding a workspace member will still collide in
### Every wave agent's brief must include
The rules an agent follows itself are
**`.claude/skills/wave/agent-rules.md`**, and a brief points there rather than
restating them. That covers base and push mechanics, the gate list, the claim,
the vacuity shapes, evidence, forbidden files, GitHub conventions, the
hand-back and disk cleanup. The list below is what this section argues for. A
new rule an agent must follow goes into that file. It does not go here, and it
does not go into the next brief.
1. **Verify your base commit first.** Worktrees do **not** reliably branch from
the orchestrator's HEAD — one was created three merges behind and only
noticed because it was told to look. `git log --oneline -1` against a named