Note the coder's round-by-round decomposition habit

e416fe7e8da5 · AtlantisPleb · · parent 97ae0f42cded

Note the coder's round-by-round decomposition habit

Observed on the suite-runner delegation: one brief became three
reviewed Devin rounds, each functionally tested by the coder before
the next. The skill now says to wait for the final verdict, not the
first child exit.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GoYpb8FEmdxVErsv7ABCYi
Co-Authored-By
Claude Fable 5 <noreply@anthropic.com>

Deploy story

What this commit did to the running system — joined from the forge receipt chain, the part a commit page elsewhere cannot show.

pushed
by user · WAL seq 328 · 2026-08-25T02:32:20.067743Z

Changed files

  • modified .agents/skills/coder-headless-delegation/SKILL.md

Diff

1 file changed, +9 -0

.agents/skills/coder-headless-delegation/SKILL.md modified +9

@@ -85,6 +85,15 @@ Tell the coder, in one message:

85 85
   with your shell tool (run `--help`, a dry-run, the tests) and report a
86 86
   verdict."
87 87
88
## Expect decomposition
89
90
Given one large brief, the coder tends to split the work into sequential
91
delegate rounds — each child gets one requirement slice, and the coder
92
functionally tests the increment itself (diffs, `--help`, dry-runs, even
93
planted-secret leak checks) before delegating the next round. This is the
94
supervision loop working; do not interrupt between rounds. The turn is over
95
when the coder prints a final verdict, not when the first child exits.
96
88 97
## Monitoring
89 98
90 99
- `tail`/`grep` `$SCRATCH/coder-session.log` for `[tool] delegate` and the

This page updates live while a promote is in flight · changelog