Correctness sub-agent: detect GitHub Actions expression evaluation pitfalls

[fullsend-ai-retro]

What happened

On PR #1611, human reviewer waynesun09 caught a HIGH bug: the ternary expression condition && '' || fallback in dispatch.yml always evaluates to fallback because empty string '' is falsy in GitHub Actions expressions. This meant the role-based kill switch was completely bypassed — when role-check set skipped=true, the expression still returned the stage name and the workflow ran. The review bot missed this across all 4 review cycles. waynesun09 also noted that fromJSON() in concurrency groups could hard-fail at workflow parse time on malformed payloads — another GHA expression pitfall the bot did not flag.

What could go better

The correctness sub-agent (skills/pr-review/sub-agents/correctness.md) has no specific guidance about GitHub Actions expression semantics. GHA expressions have non-obvious truthy/falsy rules that differ from most programming languages: empty string is falsy, 0 is falsy, and fromJSON() evaluates at parse time. The condition && '' || fallback anti-pattern is especially dangerous because it looks correct to anyone familiar with JavaScript or shell ternary idioms. No existing open issue covers GHA expression evaluation pitfalls specifically — #1275 is about tracing workflow_call chains and #1895 is about backward-incompatible input schema changes. High confidence this is actionable: the pattern is well-defined and easy to detect with specific guidance.

Proposed change

Add a 'GitHub Actions expression pitfalls' section to skills/pr-review/sub-agents/correctness.md covering: (1) Empty string '' is falsy — condition && '' || alternative always yields alternative, making it unsuitable for ternary-style expressions. The correct pattern is condition && 'empty-sentinel' || alternative or a multi-step approach. (2) fromJSON() evaluates at workflow/job parse time, not at step runtime — malformed input causes opaque parse errors, not catchable step failures. (3) 0, false, null, and '' are all falsy; only non-empty strings and non-zero numbers are truthy. Include before/after examples for each pattern.

Validation criteria

Run the correctness sub-agent against a test diff containing a workflow with the condition && '' || fallback pattern. The sub-agent should flag it as a logic bug, explaining that empty string is falsy in GHA expressions and the expression always evaluates to the fallback value.

---

Generated by retro agent from fullsend-ai/fullsend#1611

[fullsend-ai-prioritize]

RICE Priority Score: 4.05

Score breakdown

Dimension	Score	Reasoning
Reach	1.5	This affects all users who rely on the review bot to catch bugs in GitHub Actions workflow files. GHA workflows are a common part of the codebase, and the falsy-expression anti-pattern can appear in any workflow PR. Multiple teams/customers are likely affected since GHA is used platform-wide, but it only applies to the subset of PRs that touch workflow files.
Impact	1.5	The missed bug was HIGH severity — a role-based kill switch was completely bypassed because empty string is falsy in GHA expressions. Catching these patterns prevents silent security/control bypasses. This is a significant pain point: the bot failed to flag a dangerous pattern across 4 review cycles, and the fix would meaningfully improve review quality for GHA workflows.
Confidence	0.9	Very high confidence. The issue is extremely well-scoped: it identifies the exact file to modify (correctness.md), lists 3 specific patterns to document with before/after examples, provides clear validation criteria (run sub-agent against test diff), and references the concrete PR where the bug was missed. No ambiguity in scope or deliverables.
Effort	0.5	Simple, well-scoped change: add a new section to an existing markdown guidance file with 3 documented patterns and examples. No architectural changes, no multi-component coordination. The patterns are well-defined and the target file is identified. Validation requires running the sub-agent against a test diff, which is straightforward.

Formula: (1.5 x 1.5 x 0.9) / 0.5 = 4.05