opencode-skills-collection 4.0.44 → 4.0.46
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/bundled-skills/.antigravity-install-manifest.json +25 -1
- package/bundled-skills/agy-delegate/SKILL.md +168 -0
- package/bundled-skills/agy-delegate/references/dispatch-and-poll.md +140 -0
- package/bundled-skills/agy-delegate/references/multi-task-queues.md +59 -0
- package/bundled-skills/agy-delegate/references/review-and-land.md +103 -0
- package/bundled-skills/agy-delegate/references/writing-the-brief.md +124 -0
- package/bundled-skills/aider-delegate/SKILL.md +227 -0
- package/bundled-skills/aider-delegate/references/dispatch-and-poll.md +154 -0
- package/bundled-skills/aider-delegate/references/multi-task-queues.md +68 -0
- package/bundled-skills/aider-delegate/references/review-and-land.md +101 -0
- package/bundled-skills/aider-delegate/references/writing-the-brief.md +142 -0
- package/bundled-skills/babysit-pr/SKILL.md +343 -0
- package/bundled-skills/claude-delegate/SKILL.md +182 -0
- package/bundled-skills/claude-delegate/references/dispatch-and-poll.md +220 -0
- package/bundled-skills/claude-delegate/references/multi-task-queues.md +66 -0
- package/bundled-skills/claude-delegate/references/review-and-land.md +117 -0
- package/bundled-skills/claude-delegate/references/writing-the-brief.md +158 -0
- package/bundled-skills/cline-delegate/SKILL.md +140 -0
- package/bundled-skills/cline-delegate/references/dispatch-and-poll.md +142 -0
- package/bundled-skills/cline-delegate/references/multi-task-queues.md +58 -0
- package/bundled-skills/cline-delegate/references/review-and-land.md +80 -0
- package/bundled-skills/cline-delegate/references/writing-the-brief.md +129 -0
- package/bundled-skills/codex-delegate/SKILL.md +177 -0
- package/bundled-skills/codex-delegate/references/dispatch-and-poll.md +161 -0
- package/bundled-skills/codex-delegate/references/multi-task-queues.md +66 -0
- package/bundled-skills/codex-delegate/references/review-and-land.md +126 -0
- package/bundled-skills/codex-delegate/references/writing-the-brief.md +125 -0
- package/bundled-skills/commandcode-delegate/SKILL.md +188 -0
- package/bundled-skills/commandcode-delegate/references/dispatch-and-poll.md +222 -0
- package/bundled-skills/commandcode-delegate/references/multi-task-queues.md +81 -0
- package/bundled-skills/commandcode-delegate/references/review-and-land.md +144 -0
- package/bundled-skills/commandcode-delegate/references/writing-the-brief.md +147 -0
- package/bundled-skills/copilot-delegate/SKILL.md +159 -0
- package/bundled-skills/copilot-delegate/references/dispatch-and-poll.md +143 -0
- package/bundled-skills/copilot-delegate/references/multi-task-queues.md +58 -0
- package/bundled-skills/copilot-delegate/references/review-and-land.md +80 -0
- package/bundled-skills/copilot-delegate/references/writing-the-brief.md +140 -0
- package/bundled-skills/cursor-delegate/SKILL.md +170 -0
- package/bundled-skills/cursor-delegate/references/dispatch-and-poll.md +162 -0
- package/bundled-skills/cursor-delegate/references/multi-task-queues.md +59 -0
- package/bundled-skills/cursor-delegate/references/review-and-land.md +93 -0
- package/bundled-skills/cursor-delegate/references/writing-the-brief.md +126 -0
- package/bundled-skills/debate-review/SKILL.md +85 -0
- package/bundled-skills/debate-review/assets/prompts/review-debate.md +51 -0
- package/bundled-skills/debate-review/assets/prompts/review-main.md +64 -0
- package/bundled-skills/debate-review/assets/prompts/review-rebuttal.md +42 -0
- package/bundled-skills/debate-review/references/comment-format.md +63 -0
- package/bundled-skills/debate-review/references/schema.md +95 -0
- package/bundled-skills/delegate-setup/SKILL.md +215 -0
- package/bundled-skills/delegate-setup/references/schema.md +100 -0
- package/bundled-skills/delegate-setup/references/setup-dialogue.md +101 -0
- package/bundled-skills/docs/integrations/jetski-cortex.md +3 -3
- package/bundled-skills/docs/integrations/jetski-gemini-loader/README.md +1 -1
- package/bundled-skills/docs/maintainers/repo-growth-seo.md +1 -1
- package/bundled-skills/docs/maintainers/skills-update-guide.md +1 -1
- package/bundled-skills/docs/users/aas-core.md +1 -1
- package/bundled-skills/docs/users/bundles.md +1 -1
- package/bundled-skills/docs/users/claude-code-skills.md +1 -1
- package/bundled-skills/docs/users/gemini-cli-skills.md +1 -1
- package/bundled-skills/docs/users/kiro-integration.md +1 -1
- package/bundled-skills/docs/users/usage.md +3 -3
- package/bundled-skills/docs/users/visual-guide.md +4 -4
- package/bundled-skills/falsify/SKILL.md +263 -0
- package/bundled-skills/falsify/references/bias-catalog.md +42 -0
- package/bundled-skills/falsify/references/mental-models.md +135 -0
- package/bundled-skills/falsify/templates/thinking-ledger.md +104 -0
- package/bundled-skills/graceful-shutdown/SKILL.md +379 -0
- package/bundled-skills/grok-delegate/SKILL.md +179 -0
- package/bundled-skills/grok-delegate/references/dispatch-and-poll.md +171 -0
- package/bundled-skills/grok-delegate/references/multi-task-queues.md +67 -0
- package/bundled-skills/grok-delegate/references/review-and-land.md +130 -0
- package/bundled-skills/grok-delegate/references/writing-the-brief.md +119 -0
- package/bundled-skills/kimi-delegate/SKILL.md +137 -0
- package/bundled-skills/kimi-delegate/references/dispatch-and-poll.md +134 -0
- package/bundled-skills/kimi-delegate/references/multi-task-queues.md +58 -0
- package/bundled-skills/kimi-delegate/references/review-and-land.md +92 -0
- package/bundled-skills/kimi-delegate/references/writing-the-brief.md +132 -0
- package/bundled-skills/omp-delegate/SKILL.md +186 -0
- package/bundled-skills/omp-delegate/references/dispatch-and-poll.md +155 -0
- package/bundled-skills/omp-delegate/references/multi-task-queues.md +59 -0
- package/bundled-skills/omp-delegate/references/review-and-land.md +95 -0
- package/bundled-skills/omp-delegate/references/writing-the-brief.md +149 -0
- package/bundled-skills/opencode-delegate/SKILL.md +189 -0
- package/bundled-skills/opencode-delegate/references/dispatch-and-poll.md +156 -0
- package/bundled-skills/opencode-delegate/references/multi-task-queues.md +68 -0
- package/bundled-skills/opencode-delegate/references/review-and-land.md +125 -0
- package/bundled-skills/opencode-delegate/references/writing-the-brief.md +142 -0
- package/bundled-skills/pi-delegate/SKILL.md +148 -0
- package/bundled-skills/pi-delegate/references/dispatch-and-poll.md +146 -0
- package/bundled-skills/pi-delegate/references/multi-task-queues.md +59 -0
- package/bundled-skills/pi-delegate/references/review-and-land.md +95 -0
- package/bundled-skills/pi-delegate/references/writing-the-brief.md +132 -0
- package/bundled-skills/qoder-delegate/SKILL.md +151 -0
- package/bundled-skills/qoder-delegate/references/dispatch-and-poll.md +119 -0
- package/bundled-skills/qoder-delegate/references/multi-task-queues.md +43 -0
- package/bundled-skills/qoder-delegate/references/review-and-land.md +66 -0
- package/bundled-skills/qoder-delegate/references/writing-the-brief.md +85 -0
- package/bundled-skills/sandbase-mcp/SKILL.md +194 -0
- package/bundled-skills/ui-slop-score/SKILL.md +2 -2
- package/bundled-skills/unsloth-finetuning/SKILL.md +352 -0
- package/bundled-skills/vibe-delegate/SKILL.md +147 -0
- package/bundled-skills/vibe-delegate/references/dispatch-and-poll.md +144 -0
- package/bundled-skills/vibe-delegate/references/multi-task-queues.md +59 -0
- package/bundled-skills/vibe-delegate/references/review-and-land.md +87 -0
- package/bundled-skills/vibe-delegate/references/writing-the-brief.md +131 -0
- package/bundled-skills/warp-delegate/SKILL.md +192 -0
- package/bundled-skills/warp-delegate/references/dispatch-and-poll.md +163 -0
- package/bundled-skills/warp-delegate/references/multi-task-queues.md +80 -0
- package/bundled-skills/warp-delegate/references/review-and-land.md +123 -0
- package/bundled-skills/warp-delegate/references/writing-the-brief.md +105 -0
- package/bundled-skills/zcode-delegate/SKILL.md +202 -0
- package/bundled-skills/zcode-delegate/references/dispatch-and-poll.md +113 -0
- package/bundled-skills/zcode-delegate/references/multi-task-queues.md +87 -0
- package/bundled-skills/zcode-delegate/references/review-and-land.md +81 -0
- package/bundled-skills/zcode-delegate/references/writing-the-brief.md +120 -0
- package/package.json +1 -1
- package/skills_index.json +660 -0
|
@@ -0,0 +1,87 @@
|
|
|
1
|
+
# Multi-task queues
|
|
2
|
+
|
|
3
|
+
A queue is a sequence of bounded tasks dispatched one at a time, each reviewed and landed before the
|
|
4
|
+
next begins. It is the shape that makes delegation worth the overhead: a migration, a mechanical
|
|
5
|
+
refactor across many files, a removal sweep.
|
|
6
|
+
|
|
7
|
+
**Run them sequentially.** Parallel dispatches against one working tree produce interleaved edits
|
|
8
|
+
that no reviewer can untangle, and `touchedFiles` stops meaning anything.
|
|
9
|
+
|
|
10
|
+
## The loop
|
|
11
|
+
|
|
12
|
+
For each task in the queue:
|
|
13
|
+
|
|
14
|
+
1. Write the brief for **this task only**.
|
|
15
|
+
2. Dispatch.
|
|
16
|
+
3. Wait for `result.json`.
|
|
17
|
+
4. Review: re-run the gates, read the diff, check scope.
|
|
18
|
+
5. Land it — a commit per task, not one commit at the end.
|
|
19
|
+
6. Note anything learned that the next brief needs.
|
|
20
|
+
|
|
21
|
+
A commit per task is what makes a queue recoverable. When task 6 of 9 goes wrong, you revert one
|
|
22
|
+
commit instead of unpicking a nine-task blob.
|
|
23
|
+
|
|
24
|
+
## Carrying constraints forward
|
|
25
|
+
|
|
26
|
+
Each brief is written cold, so constraints do not survive on their own. Keep a short block and paste
|
|
27
|
+
it into every brief in the queue:
|
|
28
|
+
|
|
29
|
+
```markdown
|
|
30
|
+
# Standing constraints (apply to every task in this queue)
|
|
31
|
+
- Do not add dependencies.
|
|
32
|
+
- Do not reformat files you are not otherwise changing.
|
|
33
|
+
- Public API in src/api/ is frozen; changing it fails review.
|
|
34
|
+
- Gates: node test/relay-smoke.mjs
|
|
35
|
+
```
|
|
36
|
+
|
|
37
|
+
Add to this block as you learn. If task 3's review caught a drive-by refactor, task 4's brief should
|
|
38
|
+
forbid it explicitly. The block is how a queue gets *more* reliable as it runs rather than less.
|
|
39
|
+
|
|
40
|
+
## Fresh session per task, usually
|
|
41
|
+
|
|
42
|
+
Start each task cold — a new dispatch with no `--session`. Independent tasks should not inherit an
|
|
43
|
+
earlier task's context, which can carry over assumptions you rejected in review.
|
|
44
|
+
|
|
45
|
+
Use `--session` **within** a task, for rework, not **between** tasks. The exception is a genuinely
|
|
46
|
+
continuous piece of work split across dispatches for length; there, continuing the session preserves
|
|
47
|
+
context that a cold brief would have to restate.
|
|
48
|
+
|
|
49
|
+
## Tracking progress
|
|
50
|
+
|
|
51
|
+
Keep a visible checklist — the queue, one line each, marked as landed. Update it after the commit,
|
|
52
|
+
not after the dispatch. A task is done when it is committed and the gates passed, not when ZCode
|
|
53
|
+
said it finished.
|
|
54
|
+
|
|
55
|
+
Record the `sessionId` of each task next to its line. If a landed task later turns out to be wrong,
|
|
56
|
+
that id is the cheapest way back into its context.
|
|
57
|
+
|
|
58
|
+
## The end-of-run coherence check
|
|
59
|
+
|
|
60
|
+
Individually correct changes can be collectively wrong. When the queue is done, review the whole
|
|
61
|
+
range as one diff:
|
|
62
|
+
|
|
63
|
+
```bash
|
|
64
|
+
git diff <commit-before-the-queue>..HEAD
|
|
65
|
+
```
|
|
66
|
+
|
|
67
|
+
Look for what per-task review structurally cannot catch:
|
|
68
|
+
|
|
69
|
+
- **Drift** — the same problem solved three different ways across tasks.
|
|
70
|
+
- **Duplication** — a helper invented in task 2 and reinvented in task 7.
|
|
71
|
+
- **Dead ends** — code added for a task that a later task made unnecessary.
|
|
72
|
+
- **Half-migrations** — the old and new patterns now both present, with no task left to finish it.
|
|
73
|
+
|
|
74
|
+
Then run the gates once more over the final state. A queue where every task passed its own gates can
|
|
75
|
+
still end with a broken tree, because task N's gates ran before task N+1 existed.
|
|
76
|
+
|
|
77
|
+
## When to stop the queue
|
|
78
|
+
|
|
79
|
+
Stop and go back to the human when:
|
|
80
|
+
|
|
81
|
+
- Two consecutive tasks need rework for the same reason — the standing constraints are wrong, not
|
|
82
|
+
the implementer.
|
|
83
|
+
- A task reveals the plan was based on a wrong assumption about the codebase.
|
|
84
|
+
- Correct completion needs a scope change. Ask; do not expand the mandate yourself.
|
|
85
|
+
|
|
86
|
+
Stopping a queue at task 4 with a clear explanation is a better outcome than finishing all nine and
|
|
87
|
+
handing over a diff you cannot defend.
|
|
@@ -0,0 +1,81 @@
|
|
|
1
|
+
# Review and land
|
|
2
|
+
|
|
3
|
+
ZCode's `result.json` contains its own summary and its own claims about gates. Treat all of it as a
|
|
4
|
+
hypothesis. You are the reviewer; the commit is yours.
|
|
5
|
+
|
|
6
|
+
## The checklist
|
|
7
|
+
|
|
8
|
+
1. **Re-run the project's gates yourself.** Every time. "Gates passed" in `finalMessage` is a claim,
|
|
9
|
+
not evidence. Run the actual commands from the brief and read their output.
|
|
10
|
+
2. **Read the diff against the brief.** Did ZCode do what was asked — nothing more, nothing less?
|
|
11
|
+
`touchedFiles` tells you where to look; `git diff` tells you what happened.
|
|
12
|
+
3. **Check the "Out of scope" list.** Was anything on it touched?
|
|
13
|
+
4. **Read the DECISIONS section** of the report. Anything ZCode chose that the brief did not specify
|
|
14
|
+
is something you now own. Surface it; do not absorb it silently.
|
|
15
|
+
5. **On a `--read-only` run**, confirm `touchedFiles` is `[]` and `readOnlyViolation` is `false`.
|
|
16
|
+
Plan mode refuses edits, but the relay measures rather than assumes — see below.
|
|
17
|
+
6. **Run your guard skills** on the diff if you have them installed. This skill produces the work;
|
|
18
|
+
those skills judge it.
|
|
19
|
+
7. For schema or migration changes, round-trip them. For removals, grep for dangling references.
|
|
20
|
+
|
|
21
|
+
## What `readOnlyViolation` means
|
|
22
|
+
|
|
23
|
+
`plan` mode refused to write in testing, but whether that refusal is enforced by ZCode's tool layer
|
|
24
|
+
or is model compliance is not established. So the relay fingerprints the repository before the run
|
|
25
|
+
and compares afterwards:
|
|
26
|
+
|
|
27
|
+
- `false` — no Git-visible change was detected. This is the expected result.
|
|
28
|
+
- `true` — something changed during a read-only run. **Stop and inspect the tree.** The relay
|
|
29
|
+
detects and reports; it does not attribute or revert.
|
|
30
|
+
- `null` — git could not report, so the tripwire has no opinion. Inspect the tree directly.
|
|
31
|
+
|
|
32
|
+
`readOnlyViolation` is `null` on `yolo` runs, where writes are the point.
|
|
33
|
+
|
|
34
|
+
## The commit boundary
|
|
35
|
+
|
|
36
|
+
**The relay never commits.** Not on success, not on a clean gate run, not ever. Committing belongs
|
|
37
|
+
to whoever reviewed the diff, and that is you.
|
|
38
|
+
|
|
39
|
+
This is not a limitation to work around. If you find yourself wanting the implementer to commit, the
|
|
40
|
+
review step is being skipped.
|
|
41
|
+
|
|
42
|
+
## Rework: the exact-session cycle
|
|
43
|
+
|
|
44
|
+
When the diff needs changes, continue the same session rather than starting cold:
|
|
45
|
+
|
|
46
|
+
```bash
|
|
47
|
+
node "<skill-dir>/scripts/relay.mjs" --brief delta-brief.txt --cd /path/to/repo --session sess_3d8fa06c-…
|
|
48
|
+
```
|
|
49
|
+
|
|
50
|
+
- Take `sessionId` from the prior `result.json`.
|
|
51
|
+
- **Send only the delta** — what was wrong and what to do instead. ZCode still has the earlier turn;
|
|
52
|
+
re-sending the original brief wastes context and invites it to redo accepted work.
|
|
53
|
+
- `--resume-last` continues the latest session for `--cd`. It is scoped to that directory rather
|
|
54
|
+
than being a global "last", which makes it safer than it sounds — but `--session` is still the
|
|
55
|
+
precise choice when you have the id.
|
|
56
|
+
- Review the result again. A rework cycle gets the same scrutiny as the first pass.
|
|
57
|
+
|
|
58
|
+
## Authorization model
|
|
59
|
+
|
|
60
|
+
Delegation is something the human opts into. Once they have — "run this queue", "proceed" —
|
|
61
|
+
committing verified, gate-passing work is the agreed contract. Two limits on that mandate:
|
|
62
|
+
|
|
63
|
+
**Surface, don't absorb.** Report ZCode's design decisions, its defensible-but-unasked turns, and
|
|
64
|
+
non-blocking nitpicks. Your reviewer's summary should let the human disagree with a choice they
|
|
65
|
+
would not have made. Silently keeping such a change makes it yours.
|
|
66
|
+
|
|
67
|
+
**Stop for scope changes.** If correct completion requires going beyond the brief — a schema change
|
|
68
|
+
the task implies, a dependency the fix needs, a refactor without which the change is unsafe — ask.
|
|
69
|
+
Do not expand the mandate yourself, and do not let the implementer expand it for you.
|
|
70
|
+
|
|
71
|
+
## What to tell the human
|
|
72
|
+
|
|
73
|
+
After landing, say what actually happened:
|
|
74
|
+
|
|
75
|
+
- what changed, and why
|
|
76
|
+
- which gates you ran, and their real output
|
|
77
|
+
- what ZCode decided that the brief did not specify
|
|
78
|
+
- anything you left undone, and why
|
|
79
|
+
|
|
80
|
+
If a gate failed, say so with the output. If a step was skipped, say that. A clean report of a messy
|
|
81
|
+
run is more useful than a confident summary that does not survive contact with the diff.
|
|
@@ -0,0 +1,120 @@
|
|
|
1
|
+
# Writing the brief
|
|
2
|
+
|
|
3
|
+
ZCode starts every dispatch cold. It has no orchestrator chat history, no memory of the last task,
|
|
4
|
+
and no idea what you already ruled out. The brief is the entire context, and a vague brief is the
|
|
5
|
+
single most common cause of a diff you have to throw away.
|
|
6
|
+
|
|
7
|
+
## How the brief reaches ZCode
|
|
8
|
+
|
|
9
|
+
The relay writes your brief to `brief.md` in its run directory and passes it with ZCode's
|
|
10
|
+
`--attach` flag, alongside a fixed one-line prompt telling ZCode to follow the attached brief.
|
|
11
|
+
|
|
12
|
+
That matters in two ways:
|
|
13
|
+
|
|
14
|
+
- **The command line stops bounding the brief.** ZCode has no stdin delivery, so a brief passed as
|
|
15
|
+
prompt text would be limited by the command line (about 32 767 characters on Windows). Attaching
|
|
16
|
+
sidesteps that entirely — write the brief the task deserves. The configured model's context window
|
|
17
|
+
still applies, so this buys room, not an unlimited budget.
|
|
18
|
+
- **The brief is a document, not a chat message.** Structure it with headings. ZCode reads it as a
|
|
19
|
+
file.
|
|
20
|
+
|
|
21
|
+
## Structure
|
|
22
|
+
|
|
23
|
+
```markdown
|
|
24
|
+
# Goal
|
|
25
|
+
One or two sentences. What must be true when this is done.
|
|
26
|
+
|
|
27
|
+
# Current state
|
|
28
|
+
Where the relevant code lives, what it does today, and anything already tried and rejected.
|
|
29
|
+
Name files with paths. Do not make ZCode hunt.
|
|
30
|
+
|
|
31
|
+
# Change
|
|
32
|
+
Exactly what to do. Be specific about behaviour, not implementation, unless the implementation
|
|
33
|
+
is the point.
|
|
34
|
+
|
|
35
|
+
# Out of scope
|
|
36
|
+
What to leave alone. This is the field that prevents scope creep — spend real effort on it.
|
|
37
|
+
|
|
38
|
+
# Gates
|
|
39
|
+
The project's ACTUAL commands, copied from its CLAUDE.md / AGENTS.md / Makefile / package.json.
|
|
40
|
+
Do not invent these.
|
|
41
|
+
|
|
42
|
+
node test/relay-smoke.mjs
|
|
43
|
+
npx skills add . --list
|
|
44
|
+
|
|
45
|
+
# Constraints
|
|
46
|
+
Anything non-negotiable: dependency limits, style rules, files that must not be touched,
|
|
47
|
+
platforms that must keep working.
|
|
48
|
+
|
|
49
|
+
# Report contract
|
|
50
|
+
End your final message with:
|
|
51
|
+
- WHAT CHANGED: one line per file, with the reason
|
|
52
|
+
- GATES: the command you ran and its actual result
|
|
53
|
+
- DECISIONS: anything you chose that the brief did not specify
|
|
54
|
+
- UNFINISHED: anything you could not complete, and why
|
|
55
|
+
|
|
56
|
+
# You will not commit
|
|
57
|
+
Do not run `git commit`, `git push`, or any command that writes to `.git`.
|
|
58
|
+
The orchestrator reviews the diff and commits.
|
|
59
|
+
```
|
|
60
|
+
|
|
61
|
+
## The rules that actually matter
|
|
62
|
+
|
|
63
|
+
**Embed the real gate commands.** The most common failure is a brief that says "run the tests" to an
|
|
64
|
+
implementer with no idea what the test command is. Discover them from the repo first, then paste
|
|
65
|
+
them. If you did not verify the command yourself, do not put it in the brief.
|
|
66
|
+
|
|
67
|
+
**Spend effort on "Out of scope".** An implementer that is uncertain tends to do more, not less. A
|
|
68
|
+
brief that says only what to change invites reformatting, drive-by refactors, and dependency
|
|
69
|
+
additions. Name the things you do not want touched.
|
|
70
|
+
|
|
71
|
+
**One task per brief.** If the brief has an "and then also", split it. Queues are for sequences —
|
|
72
|
+
see [multi-task-queues.md](multi-task-queues.md).
|
|
73
|
+
|
|
74
|
+
**Ask for a report contract.** You are going to re-verify everything anyway, but a structured report
|
|
75
|
+
tells you where to look first, and the DECISIONS section is how you catch defensible-but-unasked
|
|
76
|
+
turns before they reach your commit.
|
|
77
|
+
|
|
78
|
+
**Say it will not commit.** The relay never commits, and ZCode running under `--mode yolo` has a
|
|
79
|
+
Bash tool. Being explicit costs one line.
|
|
80
|
+
|
|
81
|
+
## Tool restrictions belong on the command line, not in the brief
|
|
82
|
+
|
|
83
|
+
If the task genuinely must not touch certain tools, do not ask politely in the brief — pass
|
|
84
|
+
`--disallowed-tools "Write,Edit,Bash"`. That denylist is enforced by ZCode: the named tools are
|
|
85
|
+
absent from the session entirely. Prose in a brief is not.
|
|
86
|
+
|
|
87
|
+
There is no allowlist counterpart — ZCode has no `--allowed-tools`. You can subtract capability, not
|
|
88
|
+
enumerate it.
|
|
89
|
+
|
|
90
|
+
## For a read-only run
|
|
91
|
+
|
|
92
|
+
A `--read-only` dispatch runs in ZCode's `plan` mode, which refuses edits. Briefs for these runs
|
|
93
|
+
should ask for a deliverable **in the final message**, since no files will change:
|
|
94
|
+
|
|
95
|
+
```markdown
|
|
96
|
+
# Goal
|
|
97
|
+
Give a second opinion on the approach below. Do not change any files.
|
|
98
|
+
|
|
99
|
+
# Agreed
|
|
100
|
+
...points both sides accept...
|
|
101
|
+
|
|
102
|
+
# Contested
|
|
103
|
+
1. <point> — Position A: ... Position B: ...
|
|
104
|
+
2. ...
|
|
105
|
+
|
|
106
|
+
# Report contract
|
|
107
|
+
For each contested point, defend or concede, and say which evidence moved you.
|
|
108
|
+
```
|
|
109
|
+
|
|
110
|
+
Then verify the result: `touchedFiles` is `[]` and `readOnlyViolation` is `false`. Plan mode's
|
|
111
|
+
refusal is measured by the relay's Git tripwire, not guaranteed by a sandbox, and
|
|
112
|
+
`readOnlyViolation` is tri-state — `true` means Git-visible changes were detected, `false` means
|
|
113
|
+
none were, and `null` means the tripwire could not tell, which calls for inspecting the tree
|
|
114
|
+
yourself rather than assuming either.
|
|
115
|
+
|
|
116
|
+
## Delta briefs for rework
|
|
117
|
+
|
|
118
|
+
When you send work back with `--session <sessionId>`, ZCode still has the earlier turn. Send only
|
|
119
|
+
what changed — the correction, not the whole original brief. Repeating the full brief wastes context
|
|
120
|
+
and invites it to redo work you already accepted. See [review-and-land.md](review-and-land.md).
|
package/package.json
CHANGED