3 min read
skill-audit
Generated source view for the actual executable
engineering/skill-auditskill. The durable routing article is Security and Review Skills. Source: skills/engineering/skill-audit/SKILL.md
Runtime Source
| Field | Value |
|---|---|
| Category | engineering |
| Origin | personal |
| Slug | skill-audit |
| Source slug | skill-audit |
| Family | Security and Review Skills |
| Source | skills/engineering/skill-audit/SKILL.md |
Bundled Resources
No bundled resource files.
Description
Scan recent Cursor agent transcripts and update or create category skills based on recurring patterns, struggles, or inefficiencies. Use when asked to audit skills, improve automations, review recent sessions for skill opportunities, or when the user says "update my skills." Shared category skills only.
Skill Source
---
name: skill-audit
description: Scan recent Cursor agent transcripts and update or create category skills based on recurring patterns, struggles, or inefficiencies. Use when asked to audit skills, improve automations, review recent sessions for skill opportunities, or when the user says "update my skills." Shared category skills only.
origin: personal
source_slug: skill-audit
---
# Skill Audit
## Overview
Review recent Cursor agent transcripts to identify opportunities for new or improved shared skills. Modify the source category folders under `skills/{engineering,productivity,personal,misc}/`; never edit `~/.cursor/skills-cursor/` (reserved for Cursor internals) or project `.cursor/skills/`.
## Workflow
### 1) Locate transcripts
Cursor stores agent transcripts per-project under:
```
~/.cursor/projects/*/agent-transcripts/
```
Each transcript directory contains a `<uuid>.jsonl` file with conversation history.
- List all project transcript directories
- Filter to transcripts modified in the last 24 hours (use `find` with `-mtime -1` or `ls -lt`)
- Read the first few lines of each `.jsonl` to understand the conversation topic
### 2) Analyze for patterns
Look for:
- **Repeated workflows** the user performs across multiple sessions (e.g., always running the same git commands, always asking for the same kind of refactor)
- **Struggles or failures** where the agent had to retry, backtrack, or the user expressed frustration
- **Skills that were invoked but didn't help** enough (check if any skill files were read but the outcome was poor)
- **Domain knowledge** that keeps being re-explained (project conventions, API patterns, framework quirks)
### 3) Decide what to create or update
Be conservative. Only create/update a skill if:
- The pattern appeared in **2+ sessions** or was a significant time sink in one session
- The skill would save meaningful effort in future sessions
- No existing skill already covers it
### 4) Create or update skills
For each new skill:
- Follow the structure in `~/.cursor/skills-cursor/create-skill/SKILL.md`
- Place in the right category, e.g. `skills/engineering/<skill-name>/SKILL.md`, `skills/productivity/<skill-name>/SKILL.md`, `skills/personal/<skill-name>/SKILL.md`, or `skills/misc/<skill-name>/SKILL.md`
- Keep under 500 lines, use progressive disclosure for detail
- Write a specific description with trigger terms
For updates to existing skills:
- Read the current skill file first
- Use `skill-creator`'s no-op pruning pass before adding text: delete vague advice, duplicates, and lines whose removal would not change behavior
- Make targeted improvements (add missing steps, fix incorrect guidance, add examples)
- Don't rewrite skills that are working fine
### 5) Report
Summarize:
- How many transcripts were reviewed
- What patterns were identified
- Which skills were created or updated (with brief rationale)
- What was considered but skipped (and why)
Timeline
- 2026-07-15 | Generated a browseable source page from the actual executable skill file. Source: skills/engineering/skill-audit/SKILL.md