correcting-mistakes
Self-corrects skill and command instructions after resolving errors. Use when Claude errors during skill/command execution and finds a solution, or when user indicates a mistake was made. Finds the relevant instruction file and applies succinct fixes.
When & Why to Use This Skill
The Correcting Mistakes skill enables Claude to autonomously refine its own instruction sets and command files. By diagnosing whether an execution error stemmed from a misunderstanding or a flawed instruction, it applies precise, succinct updates to SKILL.md and command files. This ensures continuous improvement of agent capabilities and operational reliability by codifying successful workarounds into permanent, optimized prompt logic.
Use Cases
- Case 1: Automatically fixing ambiguous or incorrect file paths in skill instructions after a runtime error is resolved.
- Case 2: Updating command logic to include missing parameters or critical steps discovered during real-world task execution.
- Case 3: Refining agent behavior based on user feedback to eliminate recurring misunderstandings or stylistic inconsistencies in output.
- Case 4: Hardening agent workflows by documenting edge-case solutions directly into the skill's instruction set to prevent future failures.
| name | correcting-mistakes |
|---|---|
| description | Self-corrects skill and command instructions after resolving errors. Use when Claude errors during skill/command execution and finds a solution, or when user indicates a mistake was made. Finds the relevant instruction file and applies succinct fixes. |
Correcting Mistakes in Skills and Commands
Trigger this skill when:
- You error during execution of a skill or command and eventually work out a solution
- The user indicates you made a mistake in executing a skill or command
- Claude fails during skill or command execution due to an edge case
Workflow
Step 1: Locate the Instruction File
Find the relevant file that needs correction:
- Skills:
.claude/skills/{skill-name}/SKILL.mdor reference files - Commands:
.claude/commands/{command-name}.md
Read the file to understand the current instructions.
Step 2: Diagnose the Root Cause
Determine: Was this your misunderstanding, or an instruction issue?
Your misunderstanding (stop here, no changes needed):
- You misread or misapplied clear instructions
- The instruction was correct but you made an execution error
- Context from the conversation led you astray, not the instruction
Instruction issue (proceed to Step 3):
- The instruction was ambiguous, misleading, or incomplete
- An edge case was found that the skill or command doesn't account for
- The instruction specified an incorrect approach
- The instruction omitted a critical step or detail
Step 3: Test Before Fixing (if applicable)
If the correction involves a script, command, or tool use:
- Test the correct approach in the current session
- Verify it works as expected
- Only proceed to Step 4 after confirmation
This prevents codifying a "fix" that doesn't actually work.
Step 4: Apply the Correction
Edit the instruction file with these principles:
Write for a reader with no memory of the error:
- State the correct approach directly
- Do not reference the mistake, the error, or what was wrong before
- Do not add "Note:" or "Important:" warnings about the pitfall
- Do not explain why this is correct (unless explanation is inherently useful)
Be succinct:
- Change only what's necessary
- Preserve the existing style and structure
- Avoid adding defensive caveats or extra context
Example - Good correction:
# Before (incorrect)
Use `synthesis.md` as the template.
# After (correct)
Use `./templates/research-synthesis.md` as the template.
Example - Bad correction (too verbose, references error):
# After (bad - references the error)
Use `./templates/research-synthesis.md` as the template.
Note: The file is named research-synthesis.md, not synthesis.md.
Step 5: Confirm
After applying the fix:
- Re-read the corrected section to verify it reads naturally
- Confirm the fix is self-contained (no orphaned references to removed content)
- Inform the user what was corrected