fix(#2869): inject FULLSEND_OUTPUT_SCHEMA into validation script env

[fullsend-ai-coder]

When a harness uses validation_loop.schema, the schema path is resolved and available at h.ValidationLoop.Schema but was not injected into the validation script's environment. Extract a validationEnv helper that conditionally adds FULLSEND_OUTPUT_SCHEMA when the schema path is set, fixing validation failures for harnesses that rely on the new field instead of env.runner.FULLSEND_OUTPUT_SCHEMA.

Co-Authored-By: Claude Opus 4.6 noreply@anthropic.com

---

Closes #2869

Post-script verification

• Branch is not main/master (fix-2869-validation-schema-env)
• Secret scan passed (gitleaks — 5638def7e6d23e4d0b3e34a721394064be855c60..HEAD)
• Pre-commit hooks passed (authoritative run on runner)
• Tests ran inside sandbox

[github-actions]

Site preview

Preview: https://fb73dffa-site.fullsend-ai.workers.dev

Commit: 1c2514eb8701ff10a14ddbcb39790f19641e5294

[codecov]

Codecov Report

❌ Patch coverage is 88.88889% with 1 line in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
internal/cli/run.go	88.88%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!

[waynesun09]

/review

[qodo-code-review]

PR Reviewer Guide 🔍

Warning

/review is deprecated. Use /agentic_review instead (removal date not yet scheduled).

Here are some key observations to aid the review process:

🎫 Ticket compliance analysis 🔶 2869 - Partially compliant Compliant requirements: Add FULLSEND_OUTPUT_SCHEMA to the runner-side validation script environment when h.ValidationLoop.Schema is set Build validation env from h.RunnerEnv + TARGET_REPO_DIR + FULLSEND_RUN_DIR, conditionally adding schema Apply the fix at the valCmd.Env construction point for runner-side validation execution Prevent validation failures for harnesses using validation_loop.schema Non-compliant requirements: (empty) Requires further human verification: Verify end-to-end that a URL-sourced harness with validation_loop.schema successfully validates on a real workflow run (runner-side validation reading $FULLSEND_OUTPUT_SCHEMA)

⏱️ Estimated effort to review: 2 🔵🔵⚪⚪⚪

🧪 PR contains tests

🔒 No security concerns identified

⚡ Recommended focus areas for review Env Precedence valCmd.Env is constructed as append(os.Environ(), validationEnv(...)), so any overlapping keys from validationEnv should override the process environment. Confirm this ordering is intentional and matches expectations for TARGET_REPO_DIR, FULLSEND_RUN_DIR, and FULLSEND_OUTPUT_SCHEMA when those might already exist in the runner environment. valCmd.Env = append(os.Environ(), validationEnv(h, hostRepositoryDir, runDir)...) valOut, valErr := valCmd.CombinedOutput() Nil/Empty Handling validationEnv conditionally adds FULLSEND_OUTPUT_SCHEMA only when h.ValidationLoop != nil and Schema != "". Validate that all harness-loading paths consistently populate h.ValidationLoop.Schema with a host-accessible path (not sandbox path) and that no other validation-time code expects the variable to exist when Schema is omitted. func validationEnv(h *harness.Harness, hostRepoDir, runDir string) []string { env := append(envToList(h.RunnerEnv), fmt.Sprintf("TARGET_REPO_DIR=%s", hostRepoDir), fmt.Sprintf("FULLSEND_RUN_DIR=%s", runDir), ) if h.ValidationLoop != nil && h.ValidationLoop.Schema != "" { env = append(env, fmt.Sprintf("FULLSEND_OUTPUT_SCHEMA=%s", h.ValidationLoop.Schema)) } return env

[waynesun09]

/fs-fix #2870 (comment) and #2870 (comment)

[fullsend-ai-coder]

🤖 Fix · ⚠️ Cancelled · Started 9:42 PM UTC · Ended 9:47 PM UTC
Commit: 0123b0b · View workflow run →

[github-actions]

E2E tests did not run

E2E tests run automatically for org/repo members and collaborators on pull requests.

For other contributors, a maintainer must add the ok-to-test label after the latest push.

See E2E testing guide for details.

[fullsend-ai-retro]

🤖 Finished Retro · ✅ Success · Started 12:44 AM UTC · Completed 12:52 AM UTC
Commit: 1c2514e · View workflow run →

[fullsend-ai-retro]

Retro: PR #2870 — inject FULLSEND_OUTPUT_SCHEMA into validation script env

Timeline

Time (UTC)	Event
Jul 1, 19:41	Issue #2869 filed by ggallen — validation_loop.schema not exposed to validation script
Jul 1, 20:43	Code agent run 28546546028 starts
Jul 1, 20:55	PR #2870 opened by code agent (~12 min, +56/-6 lines, 2 files)
Jul 1, 21:01	Review agent run 28546636271 completes
Jul 1, 21:11	waynesun09 posts review: duplicate FULLSEND_OUTPUT_SCHEMA when both RunnerEnv and ValidationLoop.Schema are set (flagged by 3/5 review agents)
Jul 1, 21:38	waynesun09 triggers /fs-fix
Jul 1, 21:42	ggallen replies — explains this is a transient state per migration plan (#2849), asks to revert any fix
Jul 1, 21:42	Fix agent cancelled
Jul 1, 23:11	ggallen force-pushes (manual adjustment)
Jul 1, 23:28	ggallen approves and adds to merge queue
Jul 1, 23:28–00:32	5 merge queue ejections — TestAdminInstallUninstall fails from GitHub API rate limiting on e2e pool orgs
Jul 2, 00:40	Merged on 6th attempt (~72 min in merge queue)

Assessment

Code agent: excellent. Produced a clean, well-structured fix in 12 minutes with a helper function extraction and 3 targeted test cases. 88.9% patch coverage.

Review agent: technically correct but lacked migration context. The duplicate env var finding was valid from a code quality standpoint, but the maintainer explained it's a planned transient state (phase 1 of a 3-phase migration). This triggered a /fs-fix cycle that was immediately cancelled — wasted tokens but minimal impact. Existing issues #1273 and #1351 cover the general need for better context incorporation; not proposing a new issue.

Merge queue: major time sink. 5 ejections over 72 minutes due to e2e rate limiting, not related to the PR's changes at all. This is the primary improvement opportunity.

Proposals filed

• e2e: TestAdminInstallUninstall should handle GitHub API rate limits gracefully (in fullsend-ai/fullsend)