preyashyadav/sessionlint run
Budgeted, model-laddered `claude -p` runs in CI, with the full RunResult (per-rung cost, model used, success-check outcome) uploaded as a JSON job artifact. Wraps `sessionlint run` — see docs/phases/PHASE-4.md Task 2.
View on GitHubTrust Signals
- Scorecard Score
- not yet scored
- Maintenance Recency
- Stale
- License
- None
Inputs
| name | description | required | default |
|---|---|---|---|
| prompt | The task prompt passed to `claude -p`. | yes | — |
| model-ladder | Comma-separated models to try in order, e.g. haiku,sonnet. | yes | — |
| success-check | Shell command whose exit code defines task success (e.g. "npm test"). Never Claude's own self-reported result — see PHASE-4.md Task 2. | yes | — |
| anthropic-api-key | Passed through as ANTHROPIC_API_KEY to the `claude` CLI. | yes | — |
| budget | Optional per-rung $ budget cap (--max-budget-usd). | no | — |
| timeout | Optional wall-clock cap per rung, in seconds. | no | — |
| permission-mode | Passed through to `claude -p --permission-mode`. Default: acceptEdits. | no | acceptEdits |
Outputs
| name | description |
|---|---|
| report-path | Path to the JSON RunResult report on the runner filesystem. |