Skip to content

fix: checkup report improvements and severity correction (#6, #10-#14, #16, #18, #19)#29

Merged
0xbeekeeper merged 4 commits intomainfrom
fix/issues-6-10-11-12-13-14-16-18-19
Apr 6, 2026
Merged

fix: checkup report improvements and severity correction (#6, #10-#14, #16, #18, #19)#29
0xbeekeeper merged 4 commits intomainfrom
fix/issues-6-10-11-12-13-14-16-18-19

Conversation

@0xbeekeeper
Copy link
Copy Markdown
Contributor

Summary

Files changed

File Changes
src/scanner/rules/trojan.ts SOCIAL_ENGINEERING severity mediumhigh
skills/agentguard/scan-rules.md Rule 24 heading MEDIUM → HIGH
skills/agentguard/SKILL.md Detection table severity + scoring rules + dimension mapping + [REQUIRED] markers
skills/agentguard/scripts/checkup-report.js Chinese auto-detect, start "" fix, headless env detection, --file flag, 3s timeout

Test plan

  • npm run build — compiles cleanly
  • npm test — all 134 tests pass
  • Verify SOCIAL_ENGINEERING severity is high in source and docs
  • Generate checkup with Chinese analysis text — report should default to Chinese
  • On Windows: verify start "" opens browser without cmd popup
  • In headless env: verify no browser open attempt

Closes #6, Closes #10, Closes #11, Closes #12, Closes #13, Closes #14, Closes #16, Closes #18, Closes #19

🤖 Generated with Claude Code

0xbeekeeper and others added 4 commits April 6, 2026 10:20
…coverage

- Auto-detect Chinese from analysis content and apply zh locale on load (#12)
- Fix Windows `start` command needing empty title to avoid cmd popup (#11, #14)
- Skip browser open for headless/bot environments (Qclaw, OpenClaw, CI) (#14)
- Increase process exit timeout from 2s to 3s for slower systems (#11)
- Mark all 7 data collection checks as [REQUIRED] in SKILL.md (#10)
- Add pre-Step-4 validation checklist to ensure all dimensions have data (#10)
- Make Step 5 terminal summary mandatory with explicit instructions (#11)
- Add dimension→check mapping so models understand the full pipeline (#10, #13)

Closes #10, Closes #11, Closes #12, Closes #13, Closes #14

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…ility (#16)

- Add --file <path> argument to read JSON from file instead of stdin
- Update SKILL.md to use Write tool + --file method (avoids Windows
  cmd.exe single-quote issue where echo '<json>' outputs literal quotes)
- Add Write to allowed-tools for temp JSON file creation
- Keep stdin pipe as fallback for backward compatibility

Closes #16

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
#18, #19)

- Clarify Dimension 1 uses deductive scoring (base 100 minus findings)
- Clarify Dimensions 2-5 use additive scoring (start at 0, add points)
- Rename column headers to "Points if PASS" / "If FAIL" for clarity
- Add scoring example for Dimension 3 (network exposure) (#19)
- Add explicit composite score formula with and without Web3 (#18)
- Add tier assignment table with exact thresholds (#18)
- Add worked example showing full calculation → tier assignment

Closes #17, Closes #18, Closes #19

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Pressure language combined with command execution instructions is a direct
attack vector, not a medium-risk informational finding.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment