Works with the AI Agents you already use
memory secret sweep
AI agents often store everything you tell them in memory notes, instruction files, and conversation logs.
$1
memory secret sweep
Example session with this skill installed
I think I accidentally pasted my Stripe secret key into a chat with Claude Code earlier. Can you check my memory and logs for any leaked tokens and help me clean them up?
- Read your context and instructions
- Compiled the memory secret sweep
I have completed a scan of your Claude Code memory, instruction files, and conversation logs. I found 1 high-confidence secret and 3 test values.
Secret Scan Results
| ID | Type | Masked Value | Location | Reason |
|---|---|---|---|---|
| 1 | Stripe Secret Key | sk_live_...5h9Z | ~/.claude/projects/vault/memory/notes.md | Found in a memory note labeled "api-access". |
| 2 | Test Key | sk_test_...v92b | tests/fixtures/stripe_mock.json | Appears in a test fixture folder; marked as sample data. |
| 3 | Placeholder | YOUR_API_KEY | .env.example | Generic placeholder string in a template file. |
I also found 12 low-entropy strings in your transcript logs that appear to be internal tool IDs (e.g., toolu_...) which were ignored.
Next steps
- Run
python3 scripts/redact.py --ids 1to mask the Stripe key in your memory notes. - Follow the guide in
references/rotation.mdto revoke the existing Stripe key and generate a new one, as it has already been exposed to the model provider. - Once confirmed, run
python3 scripts/redact.py --purge [backup_name]to permanently delete the local backup containing the plain-text secret.
Connects securely to your tools. The creator never sees your data.
About this skill
AI agents often store everything you tell them in memory notes, instruction files, and conversation logs. If you accidentally paste a password or API key into a chat, it persists in plain text across your local agent configs and history. This skill identifies those leaked credentials and redacts them at your command.
What it does
- Multi-agent scanning searches memory and logs for Claude Code, Cursor, Windsurf, Gemini CLI, and Copilot.
- Pattern detection recognizes tokens for over 30 services including GitHub, OpenAI, AWS, Stripe, and Slack.
- Contextual analysis distinguishes between live production secrets and test fixtures, demo data, or public keys.
- Secure redaction masks confirmed secrets with
[REDACTED]while maintaining local backups for recovery. - Pre-commit integration identifies secrets in staged changes or repository history to prevent leaks before they happen.
How it works
- Initiate scan by specifying the target folders or requesting a full sweep of agent transcripts and git history.
- Review findings in a generated report that categorizes potential leaks by confidence level and type without revealing the full secret.
- Confirm targets by selecting specific IDs or confidence thresholds for the redaction script to process.
- Finalize and rotate by executing the masking command and following the provided rotation guide for any compromised keys.
Frameworks & tools
Python 3.9+. Supports Claude Code, Codex, Gemini CLI, Cursor, Windsurf, and standard Git workflows.
Why this beats prompting it yourself
Generic prompts often miss secrets hidden in large JSONL log files or nested agent memory directories. This skill uses specialized Python scripts to perform high-entropy scanning and pattern matching locally, ensuring no sensitive data is sent back to the LLM during the discovery phase.
Use cases
- Cleanse agent memory after accidentally pasting a production credential into a chat session.
- Audit local agent instruction files (e.g.,
CLAUDE.md,.cursorrules) for hardcoded keys. - Prevent secrets from entering a repository by setting up a local git pre-commit hook.
- Scan multi-gigabyte conversation transcripts for historical leaks before sharing logs.
Known limitations
Masking is skipped for logs modified in the last 15 minutes to avoid write conflicts. It cannot redact secrets from Git history or active .env files, as doing so would break your environment.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Security.
Requires Python 3.9+. Does not redact secrets from git history or .env files. Recent logs (<15 mins) are skipped. Heuristic detection may miss unlabelled random strings without --entropy flag.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 10 days ago
- Software
Git