keel-secure
Skill0 starsChanged 27 days agoMITView on GitHub
Security lane router and gate. Runs VibeSec over security-relevant code to catch vulnerabilities before they ship, and wires that check into keel-v2's security tripwire. Defers to keel-v2. Use when a change touches auth, input handling, secrets, uploads, queries, or any trust boundary — or when asked to review code for vulnerabilities.
In this repository16 other files it gives its agents.
- skills/keel-v2/CLAUDE.mdCLAUDE.md
- doc-syncSkill
- fused-landingSkill
- keel-contentSkill
- keel-designSkill
- keel-devSkill
- keel-financeSkill
- keel-legalSkill
- keel-marketingSkill
- keel-opsSkill
- keel-qaSkill
- keel-researchSkill
- keel-setupSkill
- keel-v2Skill
- keel-verifySkill
- lazy-loopSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 57
Do not silently skip the security pass.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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
Keel Secure — security lane router & gate
Index + gate. Unlike the other lanes this one is also a checkpoint:
any security-relevant change routed from keel-dev passes through here before
it's called done.
Deference — keel-v2 is the arbiter
This lane implements the keel-v2 security posture; it never relaxes it. A change touching danger surfaces 3/4/9 (PII, auth, secrets) is HEAVY by definition (keel-v2 LAWS). VibeSec's findings feed the keel-v2 /review tripwire "if any danger surface is live, both named enforcement layers are present."
Dispatch table
| When the task is… | Dispatch to | Type | Source |
| Write/review code for vulnerabilities (injection, authz, secrets, SSRF, XSS) | VibeSec-Skill | skill | BehiSecc/VibeSec-Skill |
Supabase auth, RLS, views, SECURITY DEFINER, storage policies, API keys/tokens — run its Security checklist in addition to VibeSec | supabase (opt-in, pinned) | plugin | supabase/agent-skills |
Two moments, not one
Design time (G2 Spec). Security enters before tickets exist: keel-dev's
spec.md §Threat model (conventions §4) lists every trust boundary with its
STRIDE-lite threats, the LAWS invariant that answers each, the two enforcement
layers, and the slice that carries them. This lane reviews that table at G2
(the built-in /security-review is the opt-in second pass for it). A boundary with no row is a
finding.
Build time (5b). The scan over the diff, checked against that table.
Gate rule (how it plugs into keel-v2)
keel-dev(or any lane) produces a change touching a trust boundary.- Route here → run
VibeSec-Skillover the diff. - Map each finding to a keel-v2 invariant (server-enforces, identity-from- session, exact-money, immutable-records). A finding with only one enforcement layer present = PARTIAL, not done (keel-v2 LAWS).
- Walk the §Threat model rows for the slices in this diff: each mitigation must be visible in the code, both layers. A row whose slice shipped without it = finding, same severity as the threat.
- Unresolved high-severity finding = block the commit gate. Upgrading rigor is yours to enforce; downgrading is only the human's explicit call.
- Report with
keel-verifyshape: the scanner's actual output (or "manual, no scanner" when VibeSec isn't installed), never a summary that claims more than it shows.
If the target isn't installed
Do not silently skip the security pass. Say VibeSec isn't installed, run
the keel-v2 tripwire manually (trace each trust boundary, name both enforcement
layers), and flag that an automated scan was unavailable. Install:
./install.sh secure.
Sources: BehiSecc/VibeSec-Skill (vibesec.sh). Upstream unmodified.
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.