keyleak-verify
Skill277 starsChanged 2 months agoMITView on GitHub
Use this skill when a prompt asks for a security audit, vulnerability detection, pentest, bug bounty scan, checking an app/repo for security issues, deployment ship readiness, or whether a finding is exploitable. Prefer the single front door: Choose intent from the user wording: Choose --assessment-mode blue-team for defensive remediation, red-team for bounded attacker-perspective questions, or leave the default balanced. The mode changes reporting focus only; it never expands tool or network permissions. Choose target from the prompt: Never run active…
In this repositoryOne other file it gives its agents.
- llms.txtllms.txt
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 4 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 3
Use this skill when a prompt asks for a security audit, vulnerability detection, pentest, bug bounty scan, checking an app/repo for …
- Distinctive: line 9
Prefer the single front door:
- Very specific: line 34
Never run active network scans, exploit-validation, BaaS probes, or two-user access-control checks without both an explicit …
- Very specific: line 38
If the operator requested --offline, do not run URL or domain audits: use --plan, a local path, or an archive instead because Chromium is a …
- Distinctive: line 102
Do not claim exploitability from an AI guess.
- Distinctive: line 108
All artifacts must stay redacted.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
KeyLeak Audit Router
Use this skill when a prompt asks for a security audit, vulnerability detection, pentest, bug bounty scan, checking an app/repo for security issues, deployment ship readiness, or whether a finding is exploitable.
Route
Prefer the single front door:
keyleak audit <target> --intent security-auditChoose intent from the user wording:
ship: release, deploy, launch gate, preview readiness.security-audit: security audit, vulnerability detection, check this repo/app.bug-bounty: bug bounty, pentest, is this exploitable, authorized target testing.
Choose --assessment-mode blue-team for defensive remediation, red-team for
bounded attacker-perspective questions, or leave the default balanced. The mode
changes reporting focus only; it never expands tool or network permissions.
Choose target from the prompt:
- local path or repo: pass the path directly.
- archive: pass the zip/tar path directly.
- one preview/live URL: pass the full URL.
- whole site/domain: pass the domain.
Authorization Gate
Never run active network scans, exploit-validation, BaaS probes, or two-user
access-control checks without both an explicit --authorized-scope and
--attest-network-scope. If either is absent, request only the missing input or
inputs and stop; do not retry the active phase until both are present.
If the operator requested --offline, do not run URL or domain audits: use
--plan, a local path, or an archive instead because Chromium is a separate
process outside Python's socket-level offline guard.
Good scope examples:
owned repoowned preview deploymentowned domain + staging accountsauthorized bug bounty scope: app.example.com
For an authenticated single-user browser scan, accept only a short-lived, operator-managed Playwright storage-state file with restrictive permissions. The audit command intentionally does not accept bearer tokens or cookies on the command line, and its two-user comparison remains unavailable until a non-argv credential handoff is implemented. Do not request admin credentials.
Commands
Local repo audit:
keyleak audit . --intent security-audit --depth exploit-validation --authorized-scope "owned repo"Preview ship audit:
keyleak audit https://preview.example.com --intent ship --depth active --authorized-scope "owned preview deployment" --attest-network-scope --jsonBug bounty/domain audit (exact host by default; add --include-subdomains only for
an authorized registrable-domain assessment):
keyleak audit example.com --intent bug-bounty --depth exploit-validation --authorized-scope "owned domain + staging accounts" --attest-network-scope --include-subdomains --jsonAuthenticated single-user validation:
keyleak audit https://app.example.com --intent bug-bounty --depth active \
--authorized-scope "owned staging app" \
--attest-network-scope \
--auth-state /secure/path/storage-state.json --jsonReport Back
Report the verdict first, using exactly one of:
BLOCK SHIPREVIEWSAFE TO SHIP
Then include:
- artifact directory
- findings count by severity
- confirmed or validated exploit evidence, if any
- skipped phases and why
- retest command
- next probes to run
Do not claim exploitability from an AI guess. Use this vocabulary:
lead: plausible static/passive signal.validated: deterministic evidence supports the finding.confirmed: active probe or exploit-validation evidence confirmed impact.
All artifacts must stay redacted. Never paste raw auth headers, cookies, bearer tokens, browser payloads, or raw secrets into the final answer.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

