Cascade-Safe Edit Guard
Catches cascade failures that partial checks miss — pre-check names the risk before editing, post-check demands evidence values after, then auto-repairs in a loop while rereading the entire file every cycle until zero errors remain.
Rating
No Ratings
Sold
1
How to use
Download
What this does
Most AI agents verify only the lines they changed. This skill verifies the entire file after every fix — because the thing that breaks is never where you looked.
Pre-check → Edit → Post-check → Auto-repair loop with full-file reread.
The problem it solves
You fix line 12. Line 47 silently breaks. The agent says "done." You find out later.
Partial rereads miss cascade failures. This skill catches them by rereading the whole file every repair cycle — not just the modified section.
How it works
- Pre-check — names the single most likely failure point before touching anything. "No issues" is forbidden.
- Edit — executes only what was asked. Nothing more.
- Post-check — produces evidence values from the actual file (line numbers, key names, counts). "Confirmed" is not evidence.
- Error detection — lists every error by type and location.
- Auto-repair loop — fixes errors, then rereads the entire file. Repeats until zero errors remain.
- Clean declaration — only issued after a full-file read returns zero errors.
Loop exit conditions
- Zero errors detected → clean
- Same error appears twice in a row → stops and reports as unresolvable
- More than 5 cycles → stops and lists remaining issues
Output compression
Audit lines are kept short (≤10 words). Evidence values and findings are never compressed.
Disclaimer
This skill provides behavioral instructions for AI agents. Results depend on your agent client and model. No warranties expressed or implied.
