docs: ROADMAP from issues + gh CLI workflow; parsing diff-prefix repair (#28)

[haasonsaas]

Summary

• docs/ROADMAP.md: Backlog derived from open issues with priority (high/medium/low) and area; gh CLI workflow for triage (gh issue list --label "priority: high", gh issue edit N --add-label "..."). Shipped section notes Built-in secrets detection scanner #20 (secrets done), Robust LLM output parsing with fallback strategies #28 (parsing in progress), Verification pass to catch hallucinations #23 (verification in place).
• Parsing (issue Robust LLM output parsing with fallback strategies #28): In repair_json_candidates(), add a repair candidate that strips leading + from each line when the LLM echoes JSON with diff-style line prefixes. New test: parse_json_with_diff_prefix_artifact.
• CLAUDE.md / README: Link to docs/ and ROADMAP; Contributing points to ROADMAP and gh issue list for triage.
• GitHub labels: Created priority: high|medium|low and area: review-pipeline|plugins|platform; applied to open issues via gh issue edit.

Test plan

• cargo test parsing (69 tests including parse_json_with_diff_prefix_artifact).
• CI: version, lint, test, mutation, review.

Made with Cursor

[cursor]

PR Summary

Medium Risk
Touches core LLM response parsing; although the change is small and covered by a new test, it could alter how some borderline responses are interpreted and affect downstream review output.

Overview
Improves structured-output robustness by adding a JSON repair path in repair_json_candidates() that strips diff-style leading + prefixes when an LLM echoes JSON with patch artifacts, plus a regression test (parse_json_with_diff_prefix_artifact).

Adds docs/ROADMAP.md (issue-driven backlog + gh triage commands) and updates README.md/CLAUDE.md to point contributors at the new roadmap and high-priority issue filtering.

^{Written by Cursor Bugbot for commit ae7bd57. This will update automatically on new commits. Configure here.}

[sentry]

Bug: The logic in repair_json_candidates to strip + prefixes from JSON is unreachable because upstream functions filter out or bypass the +-prefixed content.
_{Severity: MEDIUM}

Suggested Fix

The fix should be applied in extract_json_from_code_block. The condition that checks if the content starts with [ or { should be modified to also account for content where lines might have + prefixes before the JSON structure begins.

Prompt for AI Agent

Review the code at the location below. A potential bug has been identified by an AI
agent.
Verify if this is a real issue. If it is, propose a fix; if not, explain why it's not
valid.

Location: src/parsing/llm_response.rs#L358-L369

Potential issue: The new code in `repair_json_candidates` intended to strip diff-style
`+` prefixes from JSON content is unreachable. The upstream function
`extract_json_from_code_block` filters out any content that does not start with `[` or
`{`, so `+`-prefixed strings are never passed down. Fallback functions like
`find_json_array` extract valid JSON from within the prefixed lines, bypassing the new
repair logic entirely. As a result, the feature does not work as intended, and the
associated test passes due to a fallback mechanism, not because the new code is
functioning correctly.

_{Did we get this right? 👍 / 👎 to inform future reviews.}

[cursor]

Cursor Bugbot has reviewed your changes and found 2 potential issues.

^{Bugbot Autofix is ON, but it could not run because the branch was deleted or merged before autofix could start.}

[cursor]

Diff-prefix repair unreachable for multi-line JSON input

Medium Severity

The diff-prefix stripping in repair_json_candidates is unreachable for its intended multi-line use case. extract_json_from_code_block rejects code block content starting with + (the starts_with('[') || starts_with('{') guard at line 288), and find_balanced_json can't return content containing embedded + characters because serde validation fails. The repair lives downstream of extraction, so content that needs repairing never reaches it. The test parse_json_with_diff_prefix_artifact passes only because find_json_array locates the valid […] substring directly, never exercising the new repair code.

Additional Locations (2)

• src/parsing/llm_response.rs#L1234-L1244
• src/parsing/llm_response.rs#L285-L290

 

[cursor]

Repair candidates not composed, combined fixes missed

Low Severity

The diff-prefix stripping candidate is derived from trimmed (original input), and the trailing-comma candidate is also derived from trimmed. Neither repair is applied to the other's output, so if the extracted JSON needs both fixes (e.g., +-prefixed lines and trailing commas), no single candidate in the returned Vec will be valid JSON and parsing silently fails. The diff-prefix candidate also needs trailing-comma removal applied to it.