Same pattern as FORGE_URL — the llama container sets FORGE_TOKEN
to dev-qwen token via FORGE_TOKEN_OVERRIDE, but env.sh sources .env
which clobbers it back to dev-bot. All PRs and issue claims show
dev-bot instead of dev-qwen, and assignee locking fails.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Some models (especially local) emit end_turn prematurely. After
agent_run completes, check if code was pushed. If not, resume the
session with a nudge: "You stopped but did not push. Complete the
implementation, commit, and push."
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Local llama does not support claude --resume (no server-side session
storage). Stale .sid files from failed runs cause agent_run to exit
instantly on every retry, creating an infinite 1-second failure loop.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Save agent_run output to agent-run-last.json. On no_push, log the
result text, turn count, and cost. Save full output to
no-push-{issue}-{ts}.json for later analysis.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This keeps getting re-added by agents. It spins up a full Forgejo
inside CI and never finishes within the timeout.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
FORGE_TOKEN_OVERRIDE in compose env sets a per-agent token.
PRs, issue claims, and comments from the llama agent now show
dev-qwen instead of dev-bot.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The smoke-init pipeline tests `disinto init` against a Forgejo
instance — it does not build or use the agents Docker image.
Changes under docker/ should not trigger this workflow.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Use `read -rs` to hide typed secret value from terminal
- Prompt for confirmation before overwriting an existing secret
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
smoke-init spins up a full Forgejo instance inside CI and never
finishes within the 5-minute timeout. It blocks all PRs.
Remove it entirely until it can be optimized to run fast enough.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Simple while-true loop that runs dev-poll with llama backend env vars.
No cron, no guard files, no activation state — just polls and spawns.
Repo auto-cloned on first start.
To be used with a separate agents-llama compose service that sets
ANTHROPIC_BASE_URL to the llama-server address.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Delete gardener/PROMPT.md (dust-vs-ore rules already in run-gardener.toml)
- Delete supervisor/PROMPT.md (content covered by run-supervisor.toml;
migrate unique "Learning" section into formula's journal step)
- Delete vault/PROMPT.md and create formulas/run-vault.toml as the
source-of-truth formula for vault action classification/routing
- Update supervisor/supervisor-poll.sh to read from formula instead of PROMPT.md
- Update vault/vault-agent.sh to read from formula instead of PROMPT.md
- Update supervisor/AGENTS.md, vault/AGENTS.md, README.md references
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Cron does not inherit compose env vars. Without these, dev-poll fails
with cd: /home/johba/disinto: No such file or directory (host path
instead of container path).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The old skill/ reflects tmux-based pre-containerization architecture.
disinto-factory/ is the current skill with Docker Compose setup.
Closes#16
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
These 4 recipe files (cascade-rebase, chicken-egg-ci, flaky-test,
shellcheck-violations) are never referenced by any script.
The gardener uses formulas/run-gardener.toml.
Closes#23
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace tmux session discovery with .sid file globbing for stale session
cleanup and re-review triggering. Remove inject_review_into_dev_session
(dead code — both review and dev sessions now use SDK agent_run).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Prevents smoke-init and other heavy CI steps from hanging for 40+ min.
Applied automatically during disinto init.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
env.sh changes don't need a full Forgejo init smoke test.
Prevents 40-minute CI hangs on env fixes.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
source .env clobbers FORGE_URL from http://forgejo:3000 (Docker DNS)
to http://localhost:3000 (unreachable inside container). Save and
restore FORGE_URL around the source.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Root cause: env.sh skipped sourcing .env when DISINTO_CONTAINER=1,
assuming compose injects all env vars. But cron jobs do NOT inherit
compose env vars — they only get crontab-level variables.
Result: FORGE_TOKEN was empty in every cron poll. API calls returned
nothing, polls silently found "no open PRs" and exited.
Fix: always source .env regardless of DISINTO_CONTAINER. Compose env
vars (FORGE_URL) are set in the crontab env and take precedence.
Entrypoint also adds FORGE_URL to crontab env vars.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
env.sh references $USER which is not set in cron environment.
With set -u (pipefail), this causes env.sh to exit before setting
DISINTO_LOG_DIR, resulting in log writes to the read-only mount.
Root cause of silent cron failures since containerized setup.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Cron jobs run with minimal environment — no Docker compose env vars.
Without DISINTO_CONTAINER=1, env.sh falls back to FACTORY_ROOT for
log paths, which is the read-only disinto mount. Polls silently fail.
Fix: set DISINTO_CONTAINER=1 as crontab environment variable.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
