agentsclimarketplace
Sign in

Security

Skill telagod/code-abyss/skills/_kernel/security

Defensive security engineering judgment, distilled from a stronger model - invoke when THREAT MODELING a system or feature; making security-relevant design decisions (auth, crypto, trust boundaries, attack surface); reviewing code for vulnerabilities (injection, IDOR, secrets, the OWASP classes); hardening operations (secrets management, detection, patching, incident response); or judging whether a security control/program is real or theater. Also invoke BEFORE any offensive-flavored request (pentest, exploit, credential testing) to apply the authorization gate. Also invoke when designing, reviewing, or hardening an LLM/agent system that reads untrusted content or holds tool/MCP access (prompt injection, excessive agency, MCP supply chain).From its SKILL.md

Install
npx -y skills add telagod/code-abyss --skill security

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

SKILL.md

3.6 KB, 666 tokens by cl100k_base, as published. Nobody here has run it

Security — scope gate, threat model, design, review, operations, traps, agentic AI

Rule content lives in the seven files below; this SKILL.md only routes (doctrine/04-maintenance.md governs edits to this bundle too).

First, always: scope.md §1 — the authorization gate. It fires on ANY request mentioning exploits, payloads, bypassing auth/WAF/filters, cracking, brute force, credential/password testing, scanning a target, or accessing a system the requester didn't build in this session — recognize the moment even when the ask is phrased as "just write a script." Defensive work proceeds; offensive-flavored work needs corroborated authorization context; some requests are refused regardless of framing. Pass the gate before applying anything else in this bundle.

Route by moment

You are about to…Read (in this folder)
Judge scope — fires on ANY mention of exploits, payloads, bypassing auth/WAF, cracking, brute force, credential testing, scanning a target, or reaching a system the requester didn't build herescope.md §1
Decide what to protect, from whom, and where to spendthreat-model.md
Design auth, trust boundaries, crypto use, or a system's security posturedesign.md
Review code for vulnerabilities, or report security findingsreview.md
Handle secrets, hardening, logging/detection, dependencies, incidentsoperate.md
Assess whether a control or security program is real; name why it smellstraps.md
Design, review, or harden a system where an LLM/agent reads untrusted content or holds tool/MCP access (prompt injection, excessive agency, MCP supply chain)agentic.md

A new system usually runs threat-model.md (four questions) → design.md → review.md before ship → operate.md in production.

Scope and neighbors

Security judgment for building and defending systems. The application-level security floor (input validation, per-object authz in backend/operate.md §5; rate limits in backend/operate.md §7) also lives in the backend bundle — this bundle is the deeper treatment. Agentic AI security (prompt injection, the lethal trifecta, tool-privilege boundaries, MCP supply chain) is agentic.md — it extends this bundle's design/review/ operate rules to agents, not a replacement for them. General verification discipline is methods/verify.md; whether to delegate → doctrine.

The stance

Attackers take the cheapest path; defenders must be honest about which door is open. Threat model before controls (threat-model.md), assume breach (scope.md §3), make the safe way the easy way (design.md §3) — and never confuse a control that exists with a control that works (traps.md meta-signal). A clean review means "nothing found where I looked," never "secure" (scope.md §5).

What ships with it: 7 files

52.6 KB alongside SKILL.md

Gives 0 of the 12 instructions most context ai engineering skills give in 666 tokens

Counted across 1,328 of the 2,349 authors here whose files we hold, read 2026-09-06

  • Dispatch a fresh subagent for each taskin 76 of 1328, across 59 files
  • Perform spec compliance review before code quality reviewin 44 of 1328, across 34 files
  • Dispatch a final code reviewer after all tasksin 38 of 1328, across 26 files
  • Answer subagent questions before allowing implementationin 36 of 1328, across 26 files
  • Use the least powerful model capable of the taskin 33 of 1328, across 26 files
  • Create a TodoWrite list for all tasksin 32 of 1328, across 22 files
  • Perform a task review after each implementationin 31 of 1328, across 24 files
  • Extract all tasks and context from the planin 29 of 1328, across 20 files
  • Provide full task text to subagentsin 28 of 1328, across 20 files
  • Use git worktrees for isolated workspacesin 25 of 1328, across 20 files
  • Specify the model explicitly when dispatching a subagentin 23 of 1328, across 18 files
  • Execute all tasks from the plan without stoppingin 21 of 1328, across 16 files

Said here and by no other author read

  • Pass the authorization gate before applying security rules
  • Threat model before implementing security controls
  • Assume breach in system design
  • Make the safe way the easy way
  • Distinguish between existing controls and effective controls
  • Review code for vulnerabilities and security findings

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.