Move TOML frontmatter construction into a shared helper in
lib/file-action-issue.sh, used by both gardener-poll.sh and
gardener-run.sh. Fixes CI duplicate-detection failure.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Filter stale escalation entries in gardener-poll.sh before passing them
to the agent session. For each escalation reply line, extract referenced
issue numbers (#NNN) and check their current state via the API. Discard
entries where all referenced issues are already closed, preventing the
gardener from creating investigation issues for resolved problems.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Restore executable bit on gardener/gardener-poll.sh (cron invokes it directly)
- Add _BLOCKED_POSTED guard to prevent duplicate diagnostic comments when
both _on_phase_change(PHASE:crashed) and the belt-and-suspenders exit
handler both call post_blocked_diagnostic()
- Update stale documentation:
- gardener-run.sh: remove "CI escalation recipes" from issue body
- AGENTS.md: update directory layout comment for gardener-poll.sh
- gardener-poll.sh: remove recipe engine description from header
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace the unreliable escalation JSONL system (supervisor/escalations-*.jsonl
consumed by gardener) with direct blocked label + diagnostic comment on the
original issue.
When a dev-agent or action-agent session fails (PHASE:failed, idle timeout,
crash, CI exhausted):
- Capture last 50 lines from tmux pane via tmux capture-pane
- Post a structured diagnostic comment on the issue (exit reason, timestamp,
PR number, tmux output)
- Label the issue "blocked" (instead of restoring "backlog")
- Remove in-progress label
Removed:
- Escalation JSONL write paths in dev-agent.sh, phase-handler.sh, dev-poll.sh,
action-agent.sh
- is_escalated() helper in dev-poll.sh
- Escalation triage (P2f section) in supervisor-poll.sh
- Escalation processing + recipe engine in gardener-poll.sh
- ci-escalation-recipes step from run-gardener.toml formula
- escalations*.jsonl from .gitignore
Added:
- post_blocked_diagnostic() shared helper in phase-handler.sh
- ensure_blocked_label_id() helper (creates label via API if not exists)
- is_blocked() helper in dev-poll.sh (replaces is_escalated)
- Blocked issues listing in supervisor/preflight.sh
Kept:
- Matrix notifications on failure (unchanged)
- CI fix counter logic (still tracks attempts)
- needs_human injection in supervisor/gardener (not escalation-related)
- Gardener grooming (gardener-agent.sh still invoked)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The context file was written before the reset block that deleted it,
making compaction re-injection a no-op for gardener sessions.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add --recipes-only flag to gardener-poll.sh to skip grooming call when
invoked by the formula's ci-escalation-recipes step (prevents double-run)
- Update formula step to pass --recipes-only
- Add lib/file-action-issue.sh to AGENTS.md shared helpers table
- Clarify TOML arg scope in gardener trigger description
- Fix log prefixes in gardener-run.sh (poll: → run:)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
CI duplicate-detection flagged shared action-issue filing pattern between
gardener-run.sh and planner-poll.sh. Extract into lib/file-action-issue.sh
and refactor both scripts to use it.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add gardener-run.sh as a thin cron wrapper that files an action issue
referencing formulas/run-gardener.toml, following the same pattern as
planner-poll.sh. The action-agent picks up the issue and executes the
gardener formula steps in an interactive Claude session.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Widen the escalation dispatch pattern from `idle_timeout*` to also match
`idle_prompt*`. When an idle_prompt escalation arrives, the gardener now
creates an investigation sub-issue with a tailored description (session
returned to prompt without writing a phase signal) instead of silently
falling through to the recipe engine.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Fixes#261
## Changes
Fixed gardener hanging forever when Claude skips phase protocol. Three changes: (1) gardener-agent.sh: replaced 999999s timeout with 7200s (2h, matching dev-agent); (2) lib/agent-session.sh: added idle-prompt detection to monitor_phase_loop — if Claude returns to the ❯ prompt for 3 consecutive polls with no phase file written, exits immediately with _MONITOR_LOOP_EXIT=idle_prompt (only fires when phase file is empty, so awaiting_ci/review waits are unaffected); (3) gardener prompt: removed 'no time limit' wording, replaced with explicit phase-write requirement.
Co-authored-by: openhands <openhands@all-hands.dev>
Reviewed-on: https://codeberg.org/johba/disinto/pulls/263
Reviewed-by: Disinto_bot <disinto_bot@noreply.codeberg.org>
Library functions need explicit session name argument — they no longer
have closure over $SESSION_NAME from the parent script.
- agent_kill_session: add $SESSION_NAME to all 11 call sites
- agent_inject_into_session: add $SESSION_NAME to all call sites in
phase-handler.sh and gardener-agent.sh
- agent_kill_session: guard against missing arg (defensive)
kill_tmux_session → agent_kill_session
inject_into_session → agent_inject_into_session
wait_for_claude_ready → agent_wait_for_claude_ready
Also restore status() function lost during #160 refactor.
Fixes dev-agent and gardener-agent crash on startup:
line 149: status: command not found
line 280: kill_tmux_session: command not found
Remove the head -10 cap from TECH_DEBT_ISSUES so Claude sees all
tech-debt issues, not just the first 10. Apply a head -50 guard on
the list passed in PROBLEMS to avoid oversized prompts while still
feeding far more than the old cap. Update the problem label to drop
"max 10 per run" text which contradicted the zero-tech-debt objective.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Replace ${CODEBERG_WEB} with inline https://codeberg.org/${CODEBERG_REPO}
to avoid unbound variable crash in gardener-poll.sh (set -euo pipefail)
- Change sub-issue title prefix from fix: to chore: since it's an
investigation task, not a code fix
- Add emoji prefix to idle_timeout matrix notification for consistency
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Bug: chicken-egg-ci create-per-file-issues was aliased to shellcheck-only
function. Added generic playbook_lint_per_file() that handles any linter
output format. Renamed action to lint-per-file.
- Bug: cascade-rebase fired retry-merge synchronously after async rebase.
Removed retry-merge and re-approve from recipe — rebase settles, CI reruns,
normal flow handles merge on subsequent cycle.
- Warning: jq calls on PR data lacked || true under set -euo pipefail. Fixed.
- Warning: playbook_rebase_pr and playbook_retrigger_ci incremented
_PB_SUB_CREATED before confirming API success. Now check HTTP status code.
- Warning: Python import tomllib fails on < 3.11. Added try/except fallback
to tomli package.
- Nit: failures_on_unchanged regex broadened to handle generic linter formats
(file.sh:line:col patterns in addition to ShellCheck's "In file line N:").
- Info: match_recipe now logs Python stderr on error instead of silently
falling back to generic recipe.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add dedup guard: skip dust entries for issues already in dust.jsonl
- Inject already-staged issue list into LLM prompt to prevent re-emission
- Guard mv after jq: only overwrite dust.jsonl if jq succeeded
- Use sort -nu for numeric dedup of issue numbers
- Compute bundle count from distinct issues, not raw entries
- Add 30-day TTL expiry for sub-threshold dust groups
- Fix inconsistent heading levels in bundle body (all ###)
- Add scope note to PROMPT.md (human docs only, not injected)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add dust/ore rule to gardener LLM prompt: trivial tech-debt (comment
fix, rename, style-only, single-line) outputs DUST: JSON instead of
promoting individually
- Parse DUST lines from LLM output, validate JSON, append to dust.jsonl
with timestamp
- After evaluation pass: check groups with 3+ items, create bundled
backlog issue, close source issues with cross-reference
- Add gardener/dust.jsonl to .gitignore
- Create gardener/PROMPT.md documenting the dust vs ore philosophy
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Move atomic mv inside gardener loop so reply is only claimed when a
matching needs_human session exists (fixes reply-loss regression)
- Delay rm of claimed file until after successful injection in both
supervisor and gardener (OOM/SIGKILL leaves file recoverable)
- Fix matrix_listener ack message: 'next poll' instead of 'next supervisor poll'
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Fix dev-agent.sh comment: gardener-poll.sh is the backup injector, not review-poll.sh
- Add renotify marker cleanup to gardener injection path
- Use atomic mv to claim reply file, preventing double-injection race between supervisor and gardener
- Add break after supervisor injection for symmetry with gardener
- Remove overly prescriptive PHASE:awaiting_ci hardcode from injection instructions
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- supervisor: skip *.done.jsonl in escalation glob (bug: wildcard matched
harb.done.jsonl producing spurious 'pending' log noise every cycle)
- supervisor: use wc -l instead of grep -c . for line counting (style nit)
- supervisor: consume gardener-esc-resolved.log via fixed() so escalation
resolutions appear in end-of-cycle supervisor reporting
- dev-poll: update all 'escalated to supervisor' log/matrix strings to
'escalated to gardener' (lines 263, 268, 344, 420)
- gardener: track _esc_total_created across all escalation entries and
write count to supervisor/gardener-esc-resolved.log after processing
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- dev-poll.sh: write escalations to per-project files
(supervisor/escalations-{PROJECT_NAME}.jsonl) and add "project" field
so each project's escalations are isolated; update is_escalated() to
read from the same per-project paths
- gardener-poll.sh: add escalation processing block that reads the
per-project escalation file, fetches CI logs via Woodpecker, and
creates per-file ShellCheck sub-issues or generic CI failure issues
labeled backlog — runs with the correct CODEBERG_API and
WOODPECKER_REPO_ID already loaded from the project TOML
- supervisor-poll.sh: remove the escalation processing block; replace
with a simple flog report counting pending escalations per project
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Adds PRIORITY_blockers_starving_factory detection: scans backlog issues'
deps, finds any that are open but not labeled backlog, and puts them
at the top of the Claude prompt as highest priority for promotion.
Previously, gardener promoted random tech-debt issues while the actual
blockers (e.g. #650, #563, #714, #743) were ignored, leaving all
backlog items permanently stuck.
Add three new supervisor checks:
- P2c: alert when dev-agent reports "no ready issues" for 6+ consecutive polls
- P3b: detect circular dependency deadlocks via DFS cycle detection
- P3c: flag backlog issues blocked by deps open >30 days
Update supervisor PROMPT.md with guidance for Claude to resolve circular deps
by reading code context, and handle stale deps by checking relevance.
Gardener prompt now forbids bidirectional deps between sibling issues and
requires ## Related (not ## Dependencies) for cross-references.
Closes#16, Closes#17
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
These docs never existed — gardener and review-pr referenced them
as if they did. AGENTS.md tree is now the single architecture
reference for all agents.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add matrix_send() to lib/env.sh and matrix_listener.sh daemon for
real-time notifications, threaded escalations, and human-in-the-loop
replies. All agents now notify via Matrix instead of openclaw.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Claude was silently skipping ambiguous issues instead of escalating.
Made output format mandatory and explicit: every issue in the list
must result in ACTION (promoted) or ESCALATE (needs decision).
codeberg_api is a bash function in the gardener script's own process,
not available to claude-p's tool execution environment. Claude was
silently failing to call it and returning CLEAN.
Switch to curl commands with $CODEBERG_TOKEN env var that claude-p
can actually execute via its bash tool.
Tech-debt→backlog promotion was only in prompt text, not in the
problem list. Claude focused on detected problems (dupes, thin issues)
and printed CLEAN, ignoring the primary mission.
Fix: explicitly list up to 10 tech-debt issues in the problem list
so claude sees them as actionable items.
Also bumped --max-turns from 10 to 30 — promoting issues requires
reading + editing + relabeling via API, needs more turns.
