chore(dev): enforce fresh-base startup guard for CORE agents

Add scripts/agent_startup.sh — a sourced bash guard that prevents
stale-base worktrees from silently replaying or creating conflict PRs.

Root cause: PR #844 was opened from a worktree whose HEAD was behind
origin/main.  The resulting replay produced a conflict-laden PR that
required manual reconciliation (see PR #847).

## Behaviour

Default (new implementation task):
  HEAD must equal origin/main
  Working tree must be clean
  → source scripts/agent_startup.sh

PR-resume task:
  CODEX_ALLOW_NON_MAIN_BASE=1 source scripts/agent_startup.sh
  origin/main must be a strict ancestor of HEAD (enforced via merge-base)

Debug-only dirty override:
  CODEX_ALLOW_DIRTY=1 source scripts/agent_startup.sh

## Steps performed by the script

1. git fetch origin --prune
2. Print HEAD / branch / origin/main / merge-base
3. Fail if dirty (unless CODEX_ALLOW_DIRTY=1)
4. Fail if HEAD != origin/main (unless CODEX_ALLOW_NON_MAIN_BASE=1)
5. In resume mode, verify origin/main is ancestor of HEAD via merge-base
6. Print diff --name-status vs origin/main
7. Verify uv exists
8. uv sync --frozen only when uv.lock is present
9. Print final git status

## Escape hatches

CODEX_ALLOW_NON_MAIN_BASE=1   PR-resume, ancestry enforced
CODEX_ALLOW_DIRTY=1            Debug/WIP sessions, documented in session log

## Validation

bash -n scripts/agent_startup.sh  → OK
git diff --check                  → OK (no trailing whitespace)
shellcheck not available locally; shebang is bash, bash -n passed

## Non-goals

Does not touch runtime, evals, reports, teaching proposals, packs,
policy, identity, recall, vault, field, algebra, or serving paths.
This commit is contained in:
Shay 2026-06-20 19:58:33 -07:00
parent 4bab163828
commit 4b51ba146d
4 changed files with 205 additions and 5 deletions

View file

@ -112,7 +112,14 @@ Learning is reviewed mutation:
## Validation
Use CLI suites:
Before starting any task, run the startup guard to ensure a fresh base:
```bash
source scripts/agent_startup.sh
# For PR-resume tasks: CODEX_ALLOW_NON_MAIN_BASE=1 source scripts/agent_startup.sh
```
Then use CLI suites to validate your work:
```bash
core test --suite smoke -q

View file

@ -15,13 +15,18 @@ generic chatbot, not an infrastructure playground.
1. **Read this file in full.**
2. **Read `AGENTS.md` in full.**
3. **Read `docs/runtime_contracts.md` in full.**
4. **Run the smoke suite:**
4. **Run the startup guard** — enforces fresh-base and clean-tree invariants:
```bash
source scripts/agent_startup.sh
```
For a PR-resume task: `CODEX_ALLOW_NON_MAIN_BASE=1 source scripts/agent_startup.sh`
5. **Run the smoke suite:**
```bash
core test --suite smoke -q
```
5. **Check for a handoff doc** — read the most recent `HANDOFF-*.md` if one
6. **Check for a handoff doc** — read the most recent `HANDOFF-*.md` if one
exists dated within the last 3 days.
6. **State your task scope** — before editing, name the module(s) and the
7. **State your task scope** — before editing, name the module(s) and the
invariant you will prove was not violated.
---

12
GROK.md
View file

@ -104,7 +104,16 @@ Never use `git reset --hard`, broad `git checkout .`, broad `git restore .`, `gi
### 2. Establish a clean, current baseline
After the tree is clean or unknown work is safely stashed:
**Run the startup guard first** — it automates steps 24 and will hard-stop if the worktree is stale:
```bash
source scripts/agent_startup.sh
```
For a new task (default, no env vars) the script requires `HEAD == origin/main` and a clean tree.
For a PR-resume task, set `CODEX_ALLOW_NON_MAIN_BASE=1`; the script then verifies `origin/main` is a strict ancestor of `HEAD`.
If you cannot source the script, perform the equivalent steps manually:
```bash
git fetch origin --prune
@ -115,6 +124,7 @@ git status --short --branch
If `main` cannot fast-forward, stop and report the exact state. Do not merge, rebase, or resolve conflicts unless explicitly instructed.
### 3. Prefer a new worktree for non-trivial implementation
For non-trivial runtime, reasoning, eval, teaching, pack, or multi-file work, create a fresh worktree from current `origin/main`:

178
scripts/agent_startup.sh Executable file
View file

@ -0,0 +1,178 @@
#!/usr/bin/env bash
# =============================================================================
# scripts/agent_startup.sh
#
# Codex / Gemini environment startup guard for CORE.
#
# PURPOSE
# Catch stale-base worktrees before they create replay/conflict PRs.
# Every new implementation task must run from HEAD == origin/main on a
# clean tree. PR-resume tasks may run from a branch, but only when
# origin/main is a proper ancestor of HEAD.
#
# USAGE
# source scripts/agent_startup.sh # default (strict)
# CODEX_ALLOW_DIRTY=1 source ... # allow dirty worktree (debug only)
# CODEX_ALLOW_NON_MAIN_BASE=1 source ... # allow HEAD ahead of origin/main
# # (PR-resume; ancestry still checked)
#
# EXIT CODES
# 0 All checks passed — safe to proceed.
# 1 Guard failed — do NOT start the task.
#
# ENVIRONMENT CONTROLS
# CODEX_ALLOW_DIRTY=1 Bypass dirty-tree check (debug/wip only).
# CODEX_ALLOW_NON_MAIN_BASE=1 Allow HEAD ahead of origin/main, but only
# when origin/main is a strict ancestor of HEAD.
#
# NON-GOALS
# This script does NOT touch runtime, evals, reports, teaching proposals,
# packs, policy, identity, recall, vault, field, algebra, or serving paths.
# =============================================================================
set -euo pipefail
# ── helpers ──────────────────────────────────────────────────────────────────
_PASS="✓"
_FAIL="✗"
_WARN="!"
_ok() { printf '%s %s\n' "$_PASS" "$*"; }
_fail() { printf '%s %s\n' "$_FAIL" "$*" >&2; }
_warn() { printf '%s %s\n' "$_WARN" "$*" >&2; }
_sep() { printf '%s\n' "────────────────────────────────────────────────────────────────────────"; }
_die() {
_fail "$*"
_fail "Aborting. Fix the issue above, then re-run this script."
# Use 'return' rather than 'exit' so sourced scripts don't kill the shell.
return 1
}
# ── header ───────────────────────────────────────────────────────────────────
_sep
printf 'CORE agent startup guard | %s\n' "$(date -u '+%Y-%m-%dT%H:%M:%SZ')"
_sep
# ── 0. repository root ────────────────────────────────────────────────────────
REPO_ROOT="$(git rev-parse --show-toplevel 2>/dev/null)" || {
_die "Not inside a git repository."
}
_ok "repo root: $REPO_ROOT"
# ── 1. fetch origin --prune ───────────────────────────────────────────────────
printf 'Fetching origin --prune …\n'
if git fetch origin --prune --quiet; then
_ok "fetch origin --prune"
else
_die "git fetch origin --prune failed. Check network / remote config."
fi
# ── 2. diagnostic: HEAD, branch, origin/main, merge-base ──────────────────────
HEAD_SHA="$(git rev-parse HEAD)"
BRANCH="$(git rev-parse --abbrev-ref HEAD)"
ORIGIN_MAIN_SHA="$(git rev-parse origin/main 2>/dev/null)" || {
_die "origin/main does not exist. Ensure origin is correctly configured."
}
MERGE_BASE="$(git merge-base HEAD origin/main 2>/dev/null)" || {
_die "Could not compute merge-base between HEAD and origin/main."
}
printf '\n'
printf ' HEAD : %s\n' "$HEAD_SHA"
printf ' branch : %s\n' "$BRANCH"
printf ' origin/main : %s\n' "$ORIGIN_MAIN_SHA"
printf ' merge-base : %s\n' "$MERGE_BASE"
printf '\n'
# ── 3. dirty-tree check ────────────────────────────────────────────────────────
DIRTY="$(git status --porcelain)"
if [ -n "$DIRTY" ]; then
if [ "${CODEX_ALLOW_DIRTY:-0}" = "1" ]; then
_warn "Dirty worktree detected (CODEX_ALLOW_DIRTY=1, continuing in debug mode)."
else
_fail "Dirty worktree detected. Uncommitted changes are present:"
printf '%s\n' "$DIRTY" >&2
_die "Set CODEX_ALLOW_DIRTY=1 only for explicit debug sessions. For a new task, start from a clean tree."
fi
else
_ok "working tree is clean"
fi
# ── 4. base guard: HEAD == origin/main (default) or ancestry check (resume) ───
if [ "$HEAD_SHA" = "$ORIGIN_MAIN_SHA" ]; then
_ok "HEAD == origin/main (fresh base, new task allowed)"
else
if [ "${CODEX_ALLOW_NON_MAIN_BASE:-0}" = "1" ]; then
# PR-resume mode: origin/main must be an ancestor of HEAD.
if [ "$MERGE_BASE" = "$ORIGIN_MAIN_SHA" ]; then
_ok "CODEX_ALLOW_NON_MAIN_BASE=1 — origin/main is ancestor of HEAD (PR-resume allowed)"
else
_fail "CODEX_ALLOW_NON_MAIN_BASE=1 is set, but origin/main is NOT an ancestor of HEAD."
_fail " HEAD : $HEAD_SHA"
_fail " origin/main : $ORIGIN_MAIN_SHA"
_fail " merge-base : $MERGE_BASE"
_die "Your branch has diverged from origin/main. Rebase onto origin/main before resuming."
fi
else
_fail "HEAD is not origin/main."
_fail " HEAD : $HEAD_SHA (branch: $BRANCH)"
_fail " origin/main : $ORIGIN_MAIN_SHA"
_fail ""
_fail "For a new implementation task, the worktree must start from origin/main."
_fail "If you are resuming a PR, set CODEX_ALLOW_NON_MAIN_BASE=1 and ensure"
_fail "origin/main is an ancestor of your branch."
_die "Stale base detected. This is the condition that caused #844 conflict PRs."
fi
fi
# ── 5. changed files vs origin/main ───────────────────────────────────────────
printf '\nChanged files vs origin/main:\n'
DIFF_FILES="$(git diff --name-status origin/main HEAD 2>/dev/null)"
if [ -z "$DIFF_FILES" ]; then
printf ' (none — HEAD is origin/main)\n'
else
printf '%s\n' "$DIFF_FILES" | sed 's/^/ /'
fi
printf '\n'
# ── 6. uv availability ────────────────────────────────────────────────────────
if command -v uv >/dev/null 2>&1; then
UV_VERSION="$(uv --version 2>/dev/null || echo 'unknown')"
_ok "uv found: $UV_VERSION"
else
_die "uv is not installed or not in PATH. Install uv before proceeding: https://docs.astral.sh/uv/"
fi
# ── 7. uv sync --frozen (only when uv.lock exists) ────────────────────────────
LOCKFILE="$REPO_ROOT/uv.lock"
if [ -f "$LOCKFILE" ]; then
printf 'Running uv sync --frozen …\n'
if uv sync --frozen; then
_ok "uv sync --frozen succeeded"
else
_die "uv sync --frozen failed. The lock file may be out of date or the environment is broken."
fi
else
_warn "uv.lock not found — skipping uv sync (expected for projects without a lock file)"
fi
# ── 8. final git status ────────────────────────────────────────────────────────
_sep
printf 'Final git status:\n'
git status --short --branch
_sep
_ok "Startup guard passed — safe to begin task."
printf '\n'