| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
The review already ran before every commit. It ran in the context that wrote the diff, which is the part that didn't work. A self-review checks the change against the author's own model of it and can't check the model, so the errors that survive are the ones that were never in the diff: a scope inherited from whoever reported the problem, a blast radius estimated instead of measured.
So the reviewer is now a subagent, on every commit including a one-line one. It gets the diff, a one-line claim of what the change does, and the ticket where one exists. It doesn't get the conversation or my reasoning, because those are the model it exists not to hold. The ticket stays in for the opposite reason: it was written before the work and is the only thing that can contradict my claim about my own diff.
Findings loop back to the same reviewer until it approves, rather than to a fresh one that can't tell an addressed finding from one that never existed. Three rounds, or a finding that recurs after being reported fixed, and it stops for a human instead. The unattended callers park the task rather than wait for an answer nobody is awake to give.
I gave the stance a substantiation floor. An agent told to attack will manufacture findings to satisfy the instruction, and a manufactured finding costs a round and teaches the author to discount the next review.
subagents.md said three separate times not to dispatch work this small, so it now carries an Isolation Override: the size heuristics assume the main thread could do the job equally well, and they lapse when its own context is what makes its answer untrustworthy. Field 2 of the Prompt Contract inverts there. Paste your context in and you hand over the very model you spawned the agent to escape.
I wrote it by running it on itself: three rounds and thirteen findings. Two were things I'd have shipped. Withholding the ticket made my own claim self-certifying, and the override reaffirmed the contract field that would have destroyed the isolation.
|
| |
|
|
| |
subagents.md assumed the Agent tool exists. A new "Pre-Dispatch Checks" section adds two gates before any spawn: Availability (no Agent capability means do the work in the main thread under the same scope and constraints the contract would enforce) and Cost (when writing the full contract costs more than the task, do it inline). Both cross-reference the existing "Don't Subagent At All" guidance rather than duplicating it.
|
|
|
Cross-cutting rule covering subagent hygiene:
- Decision table: parallel-safe (read-only investigation, independent
test failures, multi-domain research) vs sequential-with-review
(plan execution, coupled edits) vs never-parallel (concurrent writes)
vs don't-subagent-at-all (target known, work fits in ~10 tool calls).
- Prompt contract: scope / pasted context / explicit 'do NOT' list /
required output format. Missing any field yields shallow work.
- Context-pollution rule: subagents absorb noise the main thread
shouldn't carry. When one fails, dispatch a fix-agent with the
failure report — do not retry in the orchestrator.
- Review-gate cadence: after each sequential task, or every ~3 in
parallel batches.
- Anti-patterns including parallel implementation on overlapping
files, broad 'fix all the tests' prompts, timeout-tuning to mask
flakes, and letting the agent decide scope.
Added one-line cross-references to subagents.md from debug/,
review-code/, and finish-branch/ SKILL.md — the skills that most
rely on delegation.
Clean-room synthesis from NeoLabHQ/context-engineering-kit's
subagent-driven-development pattern (MIT).
|
