Files
boxmaker/tools/run-plan.sh
T
kyleandClaude Opus 5.5 31fe078e64 opencode.json: an implement agent for unattended runs; run-plan uses it
From the combined Ornith guidance (§8.2), checked against OpenCode
1.18.30's schema and permission docs (last matching rule wins). Reads
and edits outside the repository are denied, except OpenCode's own
tool-output and temporary directories; web access, questions, subagents
and doom-loop prompts are denied; edits to AGENTS.md, CLAUDE.md,
opencode.json, the design documents, specs, plans and tools/ are denied;
shell commands are allowed except push, reset, rebase, amend, switch,
checkout, find /, locate, sudo, rm -rf / and curl/wget. A denied call
continues the loop (experimental.continue_loop_on_deny, untested with
Ornith). steps = 400 is a backstop. No custom prompt: OpenCode already
reads AGENTS.md, and a prompt may replace its tool guidance.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-09-26 00:18:03 -07:00

78 lines
3.6 KiB
Bash
Executable File

#!/bin/sh
# Runs the tasks of a plan one after another, each in its own fresh OpenCode session, and stops
# at the first task that does not end cleanly. This is the owner's tool; the implementing model
# never runs it.
#
# Usage: tools/run-plan.sh docs/plans/M2a [FIRST] FIRST is a task number such as 04
# Options (environment):
# OPENCODE_MODEL model to use (default straylight/laguna-s-2.1; OpenCode's own default differs).
# The part after the provider (`ornith-1.5-35b-a3b`) names the model in each
# commit's `Co-Authored-By: <model> <agent@wntrmute.dev>` trailer and in the log.
# OPENCODE_FLAGS extra flags for `opencode run`, for example "--pure"
# Each session runs as the `implement` agent of ./opencode.json: its permissions deny edits to the
# plans and design documents, reads outside the repository, subagents and pushes.
# DRY_RUN=1 print what would be run and stop
set -eu
plan="${1:?usage: tools/run-plan.sh docs/plans/<milestone> [FIRST]}"
first="${2:-00}"
model="${OPENCODE_MODEL:-straylight/laguna-s-2.1}"
agent="${model##*/}"
trailer="Co-Authored-By: $agent <agent@wntrmute.dev>"
log=docs/implementer-log.md
milestone=$(basename "$plan")
runs=".state/runs/$milestone"
[ -d "$plan" ] || { echo "run-plan: $plan is not a directory" >&2; exit 2; }
[ -f "$log" ] || { echo "run-plan: run this from the repository root" >&2; exit 2; }
mkdir -p "$runs"
for task in "$plan"/[0-9][0-9]-*.md; do
name=$(basename "$task" .md)
number=$(echo "$name" | cut -c1-2)
[ "$number" -ge "$first" ] || continue
prompt="Read \`$task\` and do exactly that task. You are $agent: write $agent in the log row's Model column, and end the commit message with this line, exactly: $trailer"
if [ "${DRY_RUN:-0}" = 1 ]; then
echo "opencode run --agent implement -m $model ${OPENCODE_FLAGS:-} --title $milestone/$name \"$prompt\""
continue
fi
if [ -n "$(git status --porcelain)" ]; then
echo "run-plan: the working tree is not clean before $name; stopping" >&2
exit 1
fi
before=$(git rev-parse HEAD)
echo "=== $milestone/$name: started $(date '+%H:%M:%S')"
# A new `opencode run` is a new session: nothing carries over from the task before.
# shellcheck disable=SC2086
if ! opencode run --agent implement -m "$model" ${OPENCODE_FLAGS:-} --title "$milestone/$name" "$prompt" \
> "$runs/$name.log" 2>&1; then
echo "run-plan: opencode exited with an error in $name; see $runs/$name.log" >&2
exit 1
fi
echo "=== $milestone/$name: session ended $(date '+%H:%M:%S')"
# The task must have left exactly what AGENTS.md asks for: a commit carrying the trailer, a
# clean tree, and a `done` row in the log. Anything else needs a person, so stop.
if [ "$(git rev-parse HEAD)" = "$before" ]; then
echo "run-plan: $name made no commit; stopping" >&2
exit 1
fi
if [ -n "$(git status --porcelain)" ]; then
echo "run-plan: $name left uncommitted changes; stopping" >&2
exit 1
fi
if ! git log -1 --format=%B | grep -qxF "$trailer"; then
echo "run-plan: $name's commit lacks the line '$trailer'; stopping" >&2
exit 1
fi
if ! grep -Eq "^\| *$milestone/$name *\|[^|]*\| *done *\|" "$log"; then
echo "run-plan: no 'done' row for $milestone/$name in $log; stopping" >&2
grep -E "^\| *$milestone/$name *\|" "$log" >&2 || true
exit 1
fi
echo "=== $milestone/$name: done ($(git log -1 --format=%h))"
done
echo "run-plan: every task from $first on ended cleanly. Ready for review."