rudi193-cmd/read-verify
skills/read-verify/SKILL.md
Enforce Read-Before-Edit to prevent hallucinated changes
$ install --global
skillsauthnpx skillsauth add rudi193-cmd/aionic-claude-skills read-verifyInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: Apr 13, 2026, 2:12 AM82.9s1 file scanned
SKILL.md
- name:
- read-verify
- description:
- Enforce Read-Before-Edit to prevent hallucinated changes
Read-Before-Edit Enforcement
MANDATORY RULE
You MUST Read a file before Editing it. No exceptions.
Why This Matters
Editing files without reading them leads to:
- ❌ Hallucinated changes (wrong assumptions about current code)
- ❌ Merge conflicts (overwriting recent changes)
- ❌ Breaking changes (missing context)
- ❌ Wasted tokens (generating incorrect edits)
Before ANY Edit
Step 1: Check Session History
Have I called Read(filepath) this session for this exact file?
Step 2: If NOT Read Yet
REFUSE the edit and show:
READ-BEFORE-EDIT REQUIRED
You attempted to Edit without reading the file first.
REQUIRED WORKFLOW:
- Read('path/to/file') # See current state
- Analyze content # Understand what exists
- Edit('path/to/file'...) # Then modify
WHY: Prevents hallucinated edits and ensures accuracy.
BLOCKED: Read the file first, then try Edit again.
Step 3: If Already Read
Verify the read was RECENT (within last 10 messages).
If read was >10 messages ago, require fresh read:
STALE READ DETECTED
You read this file earlier, but conversation has progressed.
File may have changed. Read it again to ensure current state.
Read('path/to/file') # Refresh your view
Acceptable Workflow
CORRECT:
- User: "Update the API endpoint in server.py"
- Claude: Read('server.py')
- [Analyze current code]
- Claude: Edit('server.py', old='...', new='...')
INCORRECT:
- User: "Update the API endpoint in server.py"
- Claude: Edit('server.py', old='...', new='...') # BLOCKED - no Read!
Edge Cases
NEW FILES:
- Write() for new files: ALLOWED (no read needed)
- Edit() for new files: BLOCKED (Edit implies file exists)
PARTIAL READS:
- Read with limit=10: Counts as read
- Read with offset: Counts as read
- Even partial reads are better than none
MULTIPLE EDITS:
- One Read can cover multiple Edits
- But if Read was >10 messages ago, require fresh read
Auto-Invoke
This skill automatically activates when:
- Edit() tool called
- File path hasn't been Read this session
- Or last Read was >10 messages ago
Implementation Note
In Claude Code CLI, this is enforced via PreToolUse hook. In claude.ai web, this skill provides the enforcement.
Related Skills
rudi193-cmd/tdd-workflow
development
VerifiedTrustedCommunity
Use this skill when writing new features, fixing bugs, or refactoring code. Enforces test-driven development with 80%+ coverage including unit, integration, and E2E tests.
2SKILL.mdUpdated Apr 13, 2026
rudi193-cmd/strategic-compact
tools
VerifiedTrustedCommunity
Suggests manual context compaction at logical intervals to preserve context through task phases rather than arbitrary auto-compaction.
2SKILL.mdUpdated Apr 13, 2026
rudi193-cmd/status
tools
VerifiedTrustedCommunity
System status — atoms, edges, ratio, boot health, unpushed work, open threads
2SKILL.mdUpdated Apr 13, 2026
rudi193-cmd/startup
tools
VerifiedTrustedCommunity
Manual session startup — when boot hooks are degraded, run the bridge-open sequence manually
2SKILL.mdUpdated Apr 13, 2026