should-auto-approve
Rates a committed code change's review risk as low (can merge on an automated approval), medium or high (needs a person). A fixed decision graph asks narrow yes/no questions about the change: the changed paths answer some, the agent answers the rest by written procedures from the code, and the graph turns the answers into the rating. The same answers always give the same rating.
- Advisory and read-only. It never approves, comments on or changes a pull request, never changes a flag, and never edits the repository it checks.
- Needs a git checkout with the change committed and its merge base,
python3(3.10+, standard library only) andgit. - Output: a short report ending in a
should-auto-approveJSON block (rating,terminal,steps,values,graph, and the approvalgateswithauto_approve). - Shadow mode: told to run in shadow mode, the agent prints the report and does nothing else with it.
Command line
should-auto-approve.pyz is the whole program in one file (a Python zip application: the walk,
the graph and its questions). start, flag, answer, decide;
python3 should-auto-approve.pyz --help lists them. SKILL.md gives the order, and
references/QUESTIONS.md every question.
Settings
| Variable | Default | What it sets |
|---|---|---|
LAUNCHDARKLY_BASE_URI |
https://app.launchdarkly.com |
The LaunchDarkly instance. |
LAUNCHDARKLY_PROJECT |
default |
The project whose flags the change may name. |
SHOULD_AUTO_APPROVE_RUNS |
~/.cache/should-auto-approve/ |
Where run folders go. |
Built from the risk graph at version 0.12 (graph e6cc6fe7a229).