Add unit test files for pre-scripts (pre-code.sh, pre-fix.sh, and others)

[fullsend-ai-retro]

What happened

PR #69 added non-empty guards (if [ -n "${GITHUB_WORKSPACE:-}" ]) to the workspace fallback loops in scripts/pre-code.sh (line 140) and scripts/pre-fix.sh (line 120), plus similar guards in scripts/post-code.sh and scripts/post-fix.sh. The post-scripts have test files (post-code-test.sh, post-fix-test.sh) and are run by make script-test (via .github/workflows/script-test.yml). However, no test files exist for pre-scripts. The Makefile's script-test target runs 8 test files — all post-scripts and validate-output-schema-test.sh — but no pre-*.sh tests. This means the guard fix in pre-code.sh and pre-fix.sh has zero regression test coverage. The code agent could not have added tests even if it tried, because there was no test infrastructure for pre-scripts to add tests to.

What could go better

The code-implementation skill mandates that every behavioral change have a corresponding test change. The code agent followed the issue spec (which only required existing tests to pass) and the AGENTS.md principle of surgical changes. But the absence of pre-script test files creates a structural gap: when the code agent fixes a bug in a pre-script, it has no way to add a regression test. If someone later removes the -n guard, no test will catch the regression.

Confidence is high that the gap exists: ls scripts/pre-*-test.sh returns nothing, and the Makefile's script-test target has no pre-script entries. Confidence is medium on the practical risk: pre-scripts run input validation on the GitHub Actions runner before sandbox creation, and broken validation could allow malformed inputs to reach the agent sandbox.

Proposed change

Create test files for the pre-scripts that contain testable logic, starting with scripts/pre-code-test.sh and scripts/pre-fix-test.sh. Each should at minimum:

1. Extract the input validation logic (ISSUE_NUMBER, REPO_FULL_NAME, GITHUB_ISSUE_URL format checks) into testable functions.
2. Include a test case that verifies the script does not crash with GITHUB_WORKSPACE unset under set -u (similar to post-triage-test.sh lines 260–267).
3. Follow the existing mock pattern used in post-script tests (mock gh commands, test in isolation).

Add the new test files to the Makefile script-test target:

script-test:
	$(call run-timed,bash scripts/pre-code-test.sh)
	$(call run-timed,bash scripts/pre-fix-test.sh)
	# ... existing entries ...

The remaining pre-scripts (pre-retro.sh, pre-review.sh, pre-scribe.sh, pre-triage.sh, pre-prioritize.sh) also lack tests but have simpler logic; prioritize pre-code.sh and pre-fix.sh since they contain the most validation logic (input checks, workspace fallback, branch detection).

Validation criteria

After implementation: (1) scripts/pre-code-test.sh and scripts/pre-fix-test.sh exist and pass when run via make script-test. (2) Each test file includes at least one test case for the GITHUB_WORKSPACE-unset scenario. (3) Deliberately removing the -n "${GITHUB_WORKSPACE:-}" guard from pre-code.sh causes the corresponding test to fail. (4) The script-test.yml CI workflow passes on the PR that adds the tests.

---

Generated by retro agent from #69

[fullsend-ai-triage]

🤖 Finished Triage · ✅ Success · Started 7:21 AM UTC · Completed 7:25 AM UTC
Commit: 32bde4b · View workflow run →

[fullsend-ai-triage]

Triage Summary

Severity: Medium · Category: Feature (test infrastructure)

Problem

Pre-scripts (pre-code.sh, pre-fix.sh) contain ~180 and ~170 lines of input validation and workspace fallback logic respectively, but have zero test coverage. The Makefile script-test target runs 8 test suites — all post-scripts — but no pre-script tests. The GITHUB_WORKSPACE non-empty guard added in PR #69 and all input validation logic have no regression protection.

Recommended Approach

Create scripts/pre-code-test.sh and scripts/pre-fix-test.sh following the existing post-script test conventions (mock gh/git, test in isolation, set -u guard tests). Add both to the Makefile script-test target.

Proposed Test Cases

# 1. GITHUB_WORKSPACE-unset guard (regression test for PR #69)
#    Unset GITHUB_WORKSPACE with set -u active
#    Expected: no 'unbound variable' crash

# 2. Input validation
#    ISSUE_NUMBER='abc' → fail | ISSUE_NUMBER='42' → pass
#    REPO_FULL_NAME='invalid' → fail | REPO_FULL_NAME='owner/repo' → pass
#    GITHUB_ISSUE_URL with bad format → fail

# 3. Cross-field consistency
#    URL repo != REPO_FULL_NAME → fail
#    URL issue number != ISSUE_NUMBER → fail
#    All fields consistent → pass

Validation Criteria (from issue)

1. Test files exist and pass via make script-test
2. Each includes at least one GITHUB_WORKSPACE-unset test case
3. Removing the -n "${GITHUB_WORKSPACE:-}" guard causes the test to fail
4. CI workflow passes on the implementing PR

Scope Note

Other pre-scripts (pre-retro.sh, pre-review.sh, pre-scribe.sh, pre-triage.sh, pre-prioritize.sh) also lack tests but have simpler logic — they can be addressed in follow-up work.

---

Labels: This is a test infrastructure enhancement. The 'enhancement' label matches the feature category. Removing 'ready-for-triage' since triage is complete.

[fullsend-ai-prioritize]

RICE Priority Score: 0.8

Score breakdown

Dimension	Score	Reasoning
Reach	1	This is internal test infrastructure for the agents repo. Pre-scripts run input validation on every GitHub Actions invocation of the code and fix agents, so a regression in these scripts could indirectly affect all users of those agents. However, the direct beneficiaries are the development team gaining regression safety. Scoring 1 for a moderate number of affected users (developers plus indirect protection of agent users).
Impact	1	The issue addresses a structural test gap: the GITHUB_WORKSPACE guard from PR #69 and all input validation logic in pre-code.sh and pre-fix.sh have zero regression protection. If someone removes the guard, no test catches it, potentially causing crashes under set -u. This is a noticeable improvement to the development workflow (regression confidence) but not blocking any current functionality. Scoring 1 for medium impact.
Confidence	0.8	The issue is exceptionally well-described with specific file paths, line numbers, validation criteria, and a concrete proposed approach. The triage comment confirms the gap with specific test case proposals. The existing post-script test patterns provide clear templates. Confidence is high at 0.8.
Effort	1	Two new test files need to be created following well-established patterns from existing post-script tests. Requires extracting testable functions from pre-scripts, writing mock commands, and adding entries to the Makefile. The patterns are clear but the pre-scripts have ~170-180 lines of logic to cover. Scoring 1 for medium effort — requires understanding context and touches a few files but is well-scoped.

Formula: (1 x 1 x 0.8) / 1 = 0.8