cloudbees-io/Clean up a feature flag with Claude Code
AI-driven feature flag cleanup action. Analyzes a repository and removes a specified feature flag, replacing all call sites with the given replacement value, then opens a pull request.
View on GitHubTrust Signals
- Scorecard Score
- not yet scored
- Maintenance Recency
- Stale
- License
- None
Inputs
| name | description | required | default |
|---|---|---|---|
| mode | Dispatch keyword: analyze | apply | yes | — |
| flag-key | Flag key to remove. Single key; one-flag-per-invocation (v0.1). | yes | — |
| replacement-value | Literal to substitute at call sites. Must match the flag's declared type — boolean ('true' or 'false'), string, or number (integer or floating-point). The skill validates this against the flag's SDK declaration. | no | — |
| language | Source language of the target repo: 'go'|'js'|'ts'|'py'|'java'. When unset, the entrypoint auto-detects by manifest priority (go > ts > js > py > java). Pass explicitly in polyglot repos to override auto-detection. | no | — |
| report-path | Absolute path to the analyze-mode report. Required in apply mode; pass through ${{ steps.<analyze>.outputs.report-path }}. | no | — |
| base-branch | Upstream branch to use as the analyze base. Defaults to the remote's default branch (origin/HEAD). Pass explicitly when cleaning a flag from a non-default branch. | no | — |
| bifrost-base-url | Bifrost gateway base URL | yes | — |
| bifrost-vk | Bifrost virtual key. Sourced from a component-level secret in the calling workflow. | yes | — |
| model | Bifrost model alias to drive Claude Code with. | no | claude-4-5-sonnet-global |
Outputs
| name | description |
|---|---|
| report-path | Absolute path to the analyze-mode report.json (under $WORKSPACE/.flag-cleanup/, excluded from the PR via .git/info/exclude). |
| outcome | Canonical outcome code. See spec.md Outcome Codes table. |
| files-modified | Count of files edited (apply modes only). |