fix(#5722): inject FULLSEND_OUTPUT_SCHEMA into post-script env

[fullsend-ai-coder]

Summary

Inject FULLSEND_OUTPUT_SCHEMA from h.ValidationLoop.Schema into the post-script environment so that post-scripts (e.g. post-fix.sh calling process-fix-result.py) can locate the JSON schema without constructing a relative path. After PR #5688 introduced directory-level script caching, the schemas/ directory is no longer co-located with scripts/ in the cache tree, breaking the relative-path fallback. This adds the env var at the post-script callsite using the same pattern as validationEnv().

Related Issue

Fixes #5722

Companion: fullsend-ai/agents#526 (updates process-fix-result.py to read the env var)

Changes

• internal/cli/run.go: After childScriptEnv() returns, conditionally append FULLSEND_OUTPUT_SCHEMA from h.ValidationLoop.Schema to postCmd.Env (lines 992-997). Uses exec.Cmd last-value-wins semantics to override any stale value from the process environment.
• internal/cli/run_test.go: Add three tests verifying the post-script schema injection: schema present when ValidationLoop.Schema is set, no append when schema is empty, no append when ValidationLoop is nil.

Testing

• New unit tests pass (TestPostScriptEnv_IncludesOutputSchema, TestPostScriptEnv_NoSchemaAppendedWhenEmpty, TestPostScriptEnv_NoSchemaAppendedWhenNoValidationLoop)
• All existing related tests pass (TestValidationEnv_*, TestPostScriptRepoEnv, TestChildScriptEnv_*, TestBootstrapEnv_*)
• go vet ./internal/cli/... passes
• go build ./internal/cli/... passes

---

Closes #5722

Post-script verification

• Branch is not main/master (agent/5722-inject-schema-post-env)
• Secret scan passed (gitleaks — 8ac8468aeabcc94dc3be41d911c7fa97508211db..HEAD)
• PR body secret scan passed (gitleaks — no-git)
• Pre-commit hooks passed (authoritative run on runner)
• Tests ran inside sandbox

[fullsend-ai-review]

🤖 Finished Review · ✅ Success · Started 3:25 PM UTC · Completed 3:39 PM UTC
Commit: 1a70eb1 · View workflow run →

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[fullsend-ai-review]

Looks good to me

Previous run

Review

Findings

Low

• [test-adequacy] internal/cli/run_test.go:2163 — The three new tests replicate the inline append logic from run.go rather than exercising the actual production code path. The production logic is a 3-line conditional inside a deferred closure in runAgent, and the tests copy that conditional verbatim. The parallel validationEnv() function is already extracted as a named helper and tested directly — the same pattern should apply here.
  Remediation: Extract the post-script env construction into a named helper function (e.g. postScriptEnv) so tests can call the real function rather than duplicating the conditional logic.

• [assertion-pattern] internal/cli/run_test.go:2193 — TestPostScriptEnv_NoSchemaAppendedWhenEmpty and TestPostScriptEnv_NoSchemaAppendedWhenNoValidationLoop use a length-comparison pattern (before := len(env); assert.Equal(t, before, len(env))) to verify no entry was appended. The existing TestValidationEnv_OmitsSchema* tests use an iteration+HasPrefix pattern instead, which is both more consistent with the codebase and slightly more robust.
  Remediation: Replace the length-comparison with the iteration pattern from existing tests: for _, e := range env { assert.False(t, strings.HasPrefix(e, "FULLSEND_OUTPUT_SCHEMA="), "message") }

---

Labels: PR modifies runner lifecycle code (internal/cli/run.go) fixing a bug in post-script environment injection

[ggallen]

/fs-review

[fullsend-ai-review]

🤖 Finished Review · ✅ Success · Started 5:24 PM UTC · Completed 5:35 PM UTC
Commit: def46ec · View workflow run →

[fullsend-ai-retro]

🤖 Finished Retro · ✅ Success · Started 9:08 PM UTC · Completed 9:17 PM UTC
Commit: 08c551f · View workflow run →

[fullsend-ai-retro]

Retro: PR #5729 — inject FULLSEND_OUTPUT_SCHEMA into post-script env

Verdict: Clean workflow. No new proposals — all improvement opportunities are covered by existing open issues.

Timeline

Time (UTC)	Event
14:07–14:33	Two fix agent runs fail after PR #5688 broke process-fix-result.py's relative-path schema lookup
14:58	ggallen files issue #5722 with exact file locations, line numbers, and a ready-to-use code snippet
14:58–15:05	Triage agent runs (3 dispatched, 1 cancelled, 2 succeed). Correctly identifies root cause, severity, and adds appropriate labels (ready-to-code, priority/high, component/runner)
15:10	ggallen triggers /fs-code
15:11–15:23	Code agent produces PR #5729 in ~12 minutes — adds env var injection in run.go and 3 unit tests
15:25	rh-hemartin approves (human review, 2 min after PR creation)
15:24–15:39	Review bot approves with 2 low-severity findings: (1) tests duplicate inline logic instead of calling an extracted helper, (2) negative tests use len() comparison instead of the established for/HasPrefix iteration pattern
17:20–17:22	ggallen addresses both findings: extracts postScriptEnv helper and aligns assertion patterns. Force-pushes.
17:23–17:35	Re-review via /fs-review — review bot approves clean
21:05	Merged via merge queue

What went well

• Triage quality: Fast (3 min), accurate root-cause analysis, correct labels, concrete proposed test.
• Code agent output: Functionally correct fix in 12 minutes from a well-specified issue. The env var injection works and all three test cases are valid.
• Review bot quality: Both findings were real and actionable — the bot caught style/pattern issues that the human reviewer did not flag. The suggestions (extract helper, match assertion pattern) directly improved code quality.
• End-to-end pipeline: Issue → triage → code → review → fix → re-review → merge all worked smoothly.

Evidence for existing issues

• Code agent should align adjacent code to dominant codebase patterns when adding similar logic agents#288 (code agent should align adjacent code to dominant codebase patterns): The code agent wrote inline env var injection instead of extracting a postScriptEnv helper to match the adjacent validationEnv() pattern. The issue body referenced validationEnv() as the model pattern, but the proposed snippet was itself inline — so the code agent followed the issue's lead. The review bot caught the gap.
• Code agent should study sibling files for established patterns before generating code #2414 / Code agent: read sibling scripts to maintain pattern consistency #2030 (code agent should study sibling files for patterns): Same underlying cause — the code agent didn't read TestValidationEnv_OmitsSchema* tests to match their assertion style (for/HasPrefix iteration vs len() comparison).
• Review agent should suggest /fs-fix for auto-fixable low-severity findings to prevent manual rework loops #5001 (review agent should suggest /fs-fix for auto-fixable low-severity findings): Both review findings were mechanical and auto-fixable (extract a function, change assertion pattern). ggallen manually fixed them and force-pushed, adding ~2 hours of latency. If the review bot had suggested /fs-fix or auto-triggered a fix run, this rework could have been automated.
• Review agent should evaluate test assertion correctness, not just test presence #1748 / Review agent should flag weak test matchers that under-assert #2741 (review agent assertion evaluation): The review bot correctly identified the assertion pattern inconsistency — this is the system working as intended for these proposals.

Autonomy observations

• The review bot's findings exceeded the human reviewer's in specificity — the bot caught two pattern inconsistencies that the human approval did not flag. For this class of change (small bug fixes in internal/cli with test coverage), the review bot demonstrates strong competence.
• The human reviewer correctly determined the code was functionally correct. The bot's value-add was style/convention enforcement, not correctness.