CodeDelta GitHub Action

Measure code churn and flag AI-generated / AI-agent code on every pull request — then comment the summary on the PR, surface findings in the code-scanning (Security) tab, and optionally block the merge. Runs entirely on your own runner; no source leaves the machine and nothing phones home. Setup is two lines in a workflow file — the theoretical minimum of effort.

_{Real output: the comment above is character-for-character what CodeDelta posted on the public demo PR, running the default mode (churn + Agent Scan).}

_{The data behind the Action: inline edit = CHG_LLOC, a changed logical statement — study and reproduction hashes at codedelta.app/paper-repchurn.html.}

# .github/workflows/codedelta.ymlname: CodeDeltaon: [pull_request]permissions:
contents: readpull-requests: write # post the summary commentsecurity-events: write # upload SARIF to the code-scanning tabjobs:
codedelta:
runs-on: ubuntu-lateststeps:
- uses: actions/checkout@v4with: { fetch-depth: 0 } # full history so churn (base..head) works
- uses: code-delta-app/action@v1

That's the whole setup. Free and fully unlocked during the beta (to 31 August 2026) — no license, no secrets, no signup. The engine downloads itself. See example-workflow.yml for a fully configured workflow with gating and an AI Bill of Materials.

Setup (one minute)

1. Add the workflow above to .github/workflows/codedelta.yml. Done.

To run past the beta — or to use your own license sooner — add one secret: CODEDELTA_LICENSE = base64 of your codedelta.lic (base64 -i codedelta.lic | pbcopy), then pass it as the license input. It overrides the built-in beta license automatically. Get a license at codedelta.app.

No install on anyone's machine; it all runs on the GitHub runner.

Inputs

Input	Default	Description
license	built-in beta license	Optional. base64 of codedelta.lic, or a path to it (use a repo secret). Overrides the free beta license (valid to 31 August 2026).
engine-url	latest release	URL of the CodeDelta Linux bundle (.tar.gz). Defaults to the latest published release; override to pin a version or self-host.
path	.	Directory to scan.
mode	churn_agent	churn_agent (churn + Agent Scan — the default) / both (adds the ML AI audit) / churn (churn only) / ai_audit (AI + agent, no churn) / ai / agent.
threshold	50	AI sensitivity 0–100 (affects rating cut-offs only).
baseline	—	Path to a committed codedelta-baseline.json.
fail-on-new	false	Fail the job if there are new findings vs the baseline.
comment	true	Post the summary as a PR comment (updates in place).
sarif	true	Upload findings as SARIF to the code-scanning tab.
gate	false	Fail the job on the default AI-governance policy: egress to a non-allied jurisdiction (CN/RU/KP/IR) or the rogue exec-on-model pattern. Needs an agent scan (mode: agent/both/ai_audit/churn_agent, i.e. any mode but plain churn/ai).
gate-policy	—	Path to a custom gate policy JSON (deny_jurisdictions / allow_providers / deny_flags / max_risk). Implies gate.
bom	—	Write an AI Bill of Materials to this path.
bom-format	native	native or cyclonedx (CycloneDX 1.6).

How findings reach the PR

• Comment — a concise churn + AI summary, re-used (edited) on each push.
• Annotations / Security tab — SARIF results appear inline on the diff and in code scanning, keyed to the flagged files. (Public repos, or private repos with GitHub Advanced Security.)
• Merge gate — with baseline + fail-on-new, only files that got worse than the accepted baseline fail the check (exit code 3).

Notes

• The license is verified locally on the runner; nothing leaves the machine. AI signals are pointers for review, not verdicts.
• Generate a baseline once and commit it: codedelta-gui scan . --mode both --write-baseline codedelta-baseline.json -q
• Default mode: churn is the ~90% case — fast, pure measurement, no ML. Switch to both to add the AI audit and agent scan.