More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Credential Leak Scan

    1

    A credential in a public repository is compromised the moment it is pushed, not the moment somebody notices.

    Free

    Security scanned
    Credential Leak Scan

    Credential Leak Scan

    Example session with this skill installed

    Check this repository before I push it to a public remote. I want to know if there are any real credentials in the working tree, and I do not want the secrets printed anywhere.

    • Read your context and instructions
    • Compiled the credential leak scan

    Credential Leak Scan Report

    Scanner: secret-leak-scan v1.0.0
    Target: payments-api (working tree)
    Files scanned: 12 | Findings: 10 | Files with findings: 5
    Highest severity: critical | Rules fired: 10 of 10

    The scan read every text file in the working tree and reported 10 findings across 5 files. The other 7 files were silent, which is the point: the scanner is not tuned to shout at ordinary code.

    Findings

    RuleSeverityFileLineWhat it matched
    ENT-001criticalconfig/production.yml6Cloud access key identifier
    ENT-002criticalsrc/deploy.py4Source-hosting access token
    ENT-003criticalsrc/notifications.py3Messaging platform token
    ENT-004criticalsrc/deploy.py6Payment service secret key
    ENT-005criticalsrc/keys.py3Private key block
    ENT-006highsrc/settings.py3Connection string with an inline credential
    ENT-007highsrc/settings.py5Credential assigned to a literal value
    ENT-008mediumsrc/settings.py7High-entropy value under a credential-looking name
    ENT-009lowsrc/settings.py9Credential name still holds a placeholder value
    ENT-010lowsrc/settings.py11Credential name with an empty value

    Redacted evidence

    RuleEvidence as emitted
    ENT-001value redacted at offset 82
    ENT-002value redacted at offset 92
    ENT-003value redacted at offset 57
    ENT-004value redacted at offset 149
    ENT-005value redacted at offset 93
    ENT-006value redacted at offset 66
    ENT-007api_key = Aq3F<redacted>4Nc9
    ENT-008client_secret = Vx7Q<redacted>k0Lz (entropy 5.0 bits per character)
    ENT-009api_key = <placeholder>
    ENT-010webhook_secret = <empty>

    Severity breakdown

    SeverityFindings
    critical5
    high2
    medium1
    low2

    Recommendation

    Treat the critical and high findings as rotation items, not cleanup items for later: a credential that has already been pushed is compromised the moment it reaches the remote, and deleting the line does not undo that. The low findings are a placeholder and an empty slot, which are worth fixing so the next reviewer is not misled and the one real key in the same file is not lost in the noise. Every finding carries a rule ID, a file and a line, and no credential is printed in full, so this report can be pasted into a ticket without leaking the secret a second time.

    Next steps

    1. Rotate every critical and high finding at its issuer before you touch the tree.
    2. Read each reported file and line, then move the value to an environment variable or a secret manager.
    3. Confirm the low findings really are placeholders, then replace them with clearly invalid values.
    4. Run the scanner over the working tree before the first push, and add it to the pull request check.
    5. The scanner reads the working tree, not the commit history. If history is in scope, run a history-aware tool as well.
    6. Keep the scan in CI so a new leak is caught on the commit that introduced it.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    A credential in a public repository is compromised the moment it is pushed, not the moment somebody notices. The hard part is not finding the obvious token, it is telling a real key from a test value that looks like one, and doing it without pasting the secret into a chat or a ticket.

    This skill scans a repository or a folder for leaked credentials and reports what it found with the value redacted. It combines exact-pattern detection for known provider token shapes, Shannon entropy scoring for high-entropy strings that no pattern covers, and a placeholder recogniser so that sample values and environment variable names are not reported as incidents.

    Every finding carries a rule identifier, a severity, a file and a line. No credential value is ever printed in full. A credential assigned to a literal value keeps the first four and last four characters with the middle replaced, a known provider shape is reported only by the offset where it matched, and a placeholder or an empty slot is reported as a marker. A reviewer can confirm the match from the rule ID, the file and the line without the secret being copied anywhere. Findings are grouped into severity bands, so the output goes straight into a backlog.

    It is a static scan of the working tree. It does not read the commit history, it does not verify a credential against a live provider, and it does not promise to find everything. It is fast, deterministic, private and good at the large majority of what actually leaks.

    Offline scan only. No history scanning or live issuer verification. Pattern matching is limited to the ten provided rules. Entropy checks require credential-like names to trigger.

    How to use

    1. 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. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    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

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 4 days ago

    • Free with an Agensi account

    Requires

    Python 3.9 or newer

    Permissions

    Terminal / Shell
    Read Files

    File Scopes

    **/*

    Terminal / Shell is used to run the bundled scanner. Read Files is used to read the repository or folder you want checked. The scanner only reads: it never writes to your tree, never opens a browser and never makes a network request. It reads the working tree, not the commit history.

    Listed4 days ago

    What's inside

    Frequently Asked Questions

    No, the scanner is a read-only, offline tool. It identifies threats and generates a redacted rotation checklist, but it will never modify your source code or delete files automatically, ensuring you stay in control of the remediation process.

    The skill uses a 'redacted reporting' protocol that only displays the prefix and suffix of a secret. This prevents your AI agent's chat history from becoming a security risk itself by storing full, plaintext credentials.

    This skill is designed for any AI agent with a Python execution environment and access to standard Unix tools like grep. It is ideal for agents used in IDEs or terminal-based developer environments.

    You receive the 'secret_leak_scan.py' script, the AI agent prompt instructions, a suite of synthetic test cases for self-validation, and the structured remediation workflow templates.

    While the scanner is highly effective at identifying high-entropy strings and provider-specific patterns, it does not scan the hidden .git history directory. It focuses on the current state of your files to prioritize immediate rotation of active secrets.

    New in Security