Adds the Sepo agent backend to this repository so @sepo-agent can answer, implement, review, and fix work on GitHub. This PR adds agent infrastructure only — no application source, secrets, or branch protection are changed.
Installed from the latest stable Sepo release v0.5.0 (self-evolving/repo).
.github/ — agent workflows (28), composite actions, prompts, issue templates, and assets, merged into the existing .github/.
.gitignore — appended .agent/dist/ and .agent/node_modules/.
Preserved / skipped
All target-owned .github/ files preserved; the conflict audit found no source path that collides with an existing target path.
Root README.md, AGENTS.md, and .skills/ were not copied (install defaults).
No application source changed.
Validation
npm ci, npm run build, and npm test pass on the source at v0.5.0 (run once against the identical payload).
git diff --cached --check is clean.
Conflict audit: no source .github path collides with an existing target path.
Skipped checks
Per-repo .agent build/test not rerun; the payload is byte-identical to the validated source revision.
This PR was created by an AI agent (OpenHands) on behalf of @ashcoft.
Summary by cubic
Installs the Sepo agent backend (v0.5.0) so @sepo-agent can answer, implement, review, and fix issues and PRs in this repository. Adds the .agent/ runtime, docs, and tests, merges the .github/ agent workflows, actions, and prompts into the existing .github/, and appends .agent/dist/ and .agent/node_modules/ to .gitignore. No application source, secrets, or branch protection are changed.
Required setup after merge
Install the Sepo GitHub App on this repository (or choose another auth path from the setup guide).
Add at least one model-provider secret under Actions secrets: OPENAI_API_KEY, CLAUDE_CODE_OAUTH_TOKEN, or ANTHROPIC_API_KEY.
Run the Agent / Onboarding / Check Setup workflow and complete any remaining setup it reports.
Run the Agent / Memory / Initialization workflow if agent/memory is missing.
Confirm Issues and GitHub Actions are enabled in repository settings.
This change is
Confidence Score: 0/5
Fix the access-check gaps and publishing status before merging.
### Issue 1
.github/workflows/agent-router.yml:318-328
If `AGENT_TRIAGE_MODE=agent`, anyone who can post an issue comment containing `@sepo-agent` can start model work before their access is checked. `extract-context` accepts known associations, including `NONE`. Triage then receives the comment, provider credentials, and a write-capable GitHub token. Route access is checked only afterward.
Check access before starting triage, and give triage only the permissions it needs.
**How this was verified:** A comment from `NONE` reaches triage with credentials before `resolve-dispatch` checks route access.
### Issue 2
.agent/src/cli/resolve-dispatch.ts:139-143
If `AGENT_TRIAGE_MODE=agent` and triage returns malformed output, `normalizeDispatch` throws before `applyDispatchPolicy` checks access. This catch block then emits `route=answer` without checking whether the requester may use it. The answer job accepts that route, so a requester denied by `AGENT_ACCESS_POLICY` can still start the answer agent.
Fail closed, or pass the fallback answer through `applyDispatchPolicy` before emitting it.
**How this was verified:** A parsing error skips `applyDispatchPolicy`, emits `route=answer`, and satisfies the answer job's route condition.
### Issue 3
.github/workflows/agent-fix-pr.yml:329-330
The final status is computed before changes are pushed and reused even if publishing fails. In an orchestrated PR-fix run, a rejected push-such as when someone updates the branch during the run-still sends `SOURCE_CONCLUSION=success`. The next review then runs against the old code.
`agent-implement.yml` also announces success if its push or PR creation fails. Compute one final status from the agent, checks, push, and PR creation results. Use that status for comments and follow-up work.
---
For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.
Summary
Adds Sepo's runtime, documentation, shared actions, prompts, and 28 workflows without changing application source.
GitHub requests can now start Sepo agent workflows.
Diagram
%%{init: {'theme': 'neutral'}}%%
flowchart TD
A[GitHub mention or label] --> B[Extract request]
B --> C{Explicit route?}
C -->|No, agent triage enabled| D[Run triage with credentials]
D --> E{Parse result}
E -->|Valid| F[Check route access]
E -->|Invalid| G[Unchecked answer fallback]
C -->|Yes| F
F --> H[Answer, implement, review, or fix PR]
G --> H
H --> I[Check changes]
I --> J[Publish changes]
J --> K[Post status and choose follow-up]