luna/oh-my-opencode
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

fix(skills): register security-research builtin

Browse Source
This commit is contained in:
YeonGyu-Kim
2026-05-30 19:45:56 +09:00
parent b5e5fce236
commit 855d191bf3
8 changed files with 284 additions and 251 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/features/AGENTS.md
+4 -3
View File
@@ -16,7 +16,7 @@ File counts are NON-TEST `.ts` files only (test files co-located but excluded fr
| **background-agent** | 30 / 1 subdir (spawner/) | HIGH | yes | Task lifecycle, concurrency (5/key), 3s polling, spawner pattern, circuit breaker. Newer files: `parent-wake-notifier.ts` (587 LOC), `loop-detector`, `error-classifier`, `fallback-retry-handler`, `process-cleanup`, `subagent-spawn-limits`, `session-status-classifier`, `compaction-aware-message-resolver`. |
| **tmux-subagent** | 27 | HIGH | yes | Tmux pane management, grid planning, session orchestration via `runTmuxCommand` |
| **opencode-skill-loader** | 25 / 1 subdir (merger/) | HIGH | yes | YAML frontmatter skill discovery from 4 scopes (project > opencode > user > global) |
| **builtin-skills** | 19 / 5 subdirs | LOW–MED | yes | 11 built-in skill files (git-master, playwright, frontend-ui-ux, review-work, ai-slop-remover, security-review, dev-browser, playwright-cli, **team-mode**, …) |
| **builtin-skills** | 20 / 6 subdirs | LOW–MED | yes | 12 built-in skill files (git-master, playwright, frontend-ui-ux, review-work, ai-slop-remover, security-research, security-review, dev-browser, playwright-cli, **team-mode**, …) |
| **skill-mcp-manager** | 11 | HIGH | yes | Tier-3 MCP client lifecycle per session (stdio + HTTP + OAuth) |
| **claude-code-plugin-loader** | 11 | MEDIUM | yes | Unified Claude Code plugin discovery (commands, agents, skills, hooks, MCPs) |
| **builtin-commands** | 11 / 1 subdir (templates/) | LOW | yes | Command templates: refactor, init-deep, handoff, ulw-loop, etc. |
@@ -70,7 +70,7 @@ Eligible members: sisyphus, atlas, sisyphus-junior, hephaestus only. See [`team-
State-first tmux integration. Centralized tmux command execution through `src/shared/tmux/runner.ts` (`runTmuxCommand`). Direct `Bun.spawn(["tmux", ...])` is FORBIDDEN — would drift from retry/timeout discipline.
### builtin-skills (11 skills)
### builtin-skills (12 skills)
| Skill | LOC | MCP | Notes |
|-------|-----|-----|-------|
@@ -80,7 +80,8 @@ State-first tmux integration. Centralized tmux command execution through `src/sh
| dev-browser | 221 | — | Persistent page state browser |
| review-work | ~500 | — | 5-agent post-implementation review orchestrator |
| ai-slop-remover | ~300 | — | Remove AI code patterns |
| security-review | ~180 | — | Team Mode exploitability-driven security review |
| security-research | SKILL.md | — | Team Mode exploitability-driven security research |
| security-review | wrapper | — | Alias for security-research |
| **team-mode** | — | — | Loaded only when `team_mode.enabled` (skill explains the 12 tools to agents) |
| frontend-ui-ux | 79 | — | Design-first UI development |
| (git-master-skill-metadata) | — | — | Companion to git-master |
src/features/builtin-skills/AGENTS.md
+8 -5
View File
@@ -1,4 +1,4 @@
# src/features/builtin-skills/ — 11 Built-in Skill Files
# src/features/builtin-skills/ — 12 Built-in Skill Files
**Generated:** 2026-05-15
@@ -11,7 +11,7 @@ Skills shipped inside the plugin (always available, no install). Registered via
```
builtin-skills/
├── index.ts # Barrel exports
├── skills.ts # createBuiltinSkills() factory — registers all 11 below
├── skills.ts # createBuiltinSkills() factory — registers all 12 below
├── types.ts # BuiltinSkill interface
├── skills/
│ ├── git-master.ts # 1111 LOC
@@ -23,13 +23,15 @@ builtin-skills/
│ ├── review-work.ts # 5-agent post-implementation review
│ ├── ai-slop-remover.ts # Remove AI-generated code patterns
│ ├── team-mode.ts # 12 team_* tool documentation (gated)
│ ├── security-review.ts # Team Mode exploitability-driven security review
│ ├── security-research.ts # Team Mode exploitability-driven security research
│ ├── security-review.ts # Alias for security-research
│ ├── git-master-sections/ # Git-master prompt sub-sections
│ └── index.ts # skill barrel
├── git-master/ # Resources for git-master skill
├── frontend-ui-ux/ # Resources for frontend-ui-ux skill
├── agent-browser/ # Resources for agent-browser variant
└── dev-browser/ # Resources for dev-browser
├── dev-browser/ # Resources for dev-browser
└── security-research/ # Resources for security-research
```
## SKILL CATALOG
@@ -44,7 +46,8 @@ builtin-skills/
| `frontend-ui-ux` | 79 | — | Design-first UI development guidance |
| `review-work` | ~500 | — | Post-implementation review orchestrator (5 parallel agents) |
| `ai-slop-remover` | ~300 | — | Remove AI-generated code smells |
| `security-review` | ~180 | — | Team Mode exploitability-driven security review |
| `security-research` | SKILL.md | — | Team Mode exploitability-driven security research |
| `security-review` | wrapper | — | Alias for `security-research` |
| `team-mode` | — | — | **Conditional** — only loaded when `team_mode.enabled`; documents the 12 `team_*` tools and lifecycle |
## BROWSER VARIANT SELECTION
src/features/builtin-skills/security-research/SKILL.md
+198
View File
@@ -0,0 +1,198 @@
# Security Research - Team Mode Vulnerability Audit
Use this skill to run a parallel security audit that separates real exploitability from generic concern. The team has 3 vulnerability hunters and 2 PoC engineers.
## Hard Preconditions
Before starting, verify:
1. `team_*` tools are available. If not, stop and tell the user:
`security-research requires team-mode. Set team_mode.enabled: true in your oh-my-openagent config, restart opencode, then retry.`
2. You are in the main session, not a background subagent.
3. You have a concrete target: repository, diff range, PR, release candidate, path list, or threat surface.
If the user provided no target, audit the current repository and current branch diff against its upstream or merge base. If there is no diff, audit the security-sensitive surfaces in the working tree.
## Severity Standard
Use these references as the scoring frame:
- CWE for root-cause weakness classification: https://cwe.mitre.org/
- OWASP WSTG for test methodology: https://devguide.owasp.org/en/06-verification/01-guides/01-wstg/
- OWASP ASVS for control verification: https://owasp.org/www-project-application-security-verification-standard/
- CVSS v4.0 for exploitability and impact scoring: https://www.first.org/cvss/v4.0/specification-document
Rules:
- No severity without an attack path.
- No critical or high finding without concrete exploit preconditions and impact.
- Keep CWE category separate from severity.
- Prefer a small, reproducible PoC over theoretical language.
- Never run destructive exploits against real services or third-party systems.
- Use local fixtures, toy payloads, dry runs, or static proof when real execution would be unsafe.
## Team Roster
Create one Team Mode run with these 5 members:
| Member | Kind | Category | Role |
|--------|------|----------|------|
| `surface-hunter` | category | `deep` | Map entry points, trust boundaries, and reachable attack surfaces. |
| `auth-data-hunter` | category | `ultrabrain` | Hunt auth, authorization, data isolation, injection, and secret handling flaws. |
| `runtime-supply-hunter` | category | `unspecified-high` | Hunt filesystem, subprocess, archive, dependency, hook, MCP, and config risks. |
| `poc-engineer-a` | category | `unspecified-high` | Build minimal PoCs for the strongest candidate findings. |
| `poc-engineer-b` | category | `deep` | Independently reproduce, falsify, or downgrade candidate findings. |
Call `team_create` with an inline spec:
```typescript
team_create({
inline_spec: {
name: "security-research",
description: "Parallel exploitability-driven security research team.",
members: [
{
name: "surface-hunter",
kind: "category",
category: "deep",
prompt: "You map attack surface. Enumerate entry points, trust boundaries, attacker-controlled inputs, data sinks, privilege transitions, and sensitive assets. Return evidence with file paths and exact functions. Do not assign severity unless you can name an attack path."
},
{
name: "auth-data-hunter",
kind: "category",
category: "ultrabrain",
prompt: "You hunt auth, authorization, tenant/data isolation, injection, SSRF, credential exposure, and confused-deputy flaws. Reason from attacker capability to impact. Return only findings with concrete exploit preconditions, CWE candidates, and verification steps."
},
{
name: "runtime-supply-hunter",
kind: "category",
category: "unspecified-high",
prompt: "You hunt filesystem, subprocess, archive extraction, dependency, hook execution, MCP, config, and environment-variable risks. Check path traversal, command injection, unsafe downloads, permission boundaries, and supply-chain assumptions. Cite file paths and commands used."
},
{
name: "poc-engineer-a",
kind: "category",
category: "unspecified-high",
prompt: "You build minimal safe PoCs for candidate findings. Use toy inputs and local-only execution. Your job is to prove or disprove exploitability, not to broaden scope. Report exact reproduction steps and expected output."
},
{
name: "poc-engineer-b",
kind: "category",
category: "deep",
prompt: "You independently reproduce candidate findings and try to falsify them. Downgrade anything without a working path. If a PoC is unsafe to run, design a safe static or dry-run proof and explain the limit."
}
]
}
})
```
If a category is unavailable, retry once by replacing only that category with `unspecified-high`. Do not reduce the team below 5 members.
## Workflow
### Phase 0: Scope and Baseline
Collect:
- Target scope and reason for audit.
- Branch, base ref, diff, and changed files if this is a change review.
- Security-sensitive directories and files if this is a full-repo audit.
- Existing tests and commands that exercise relevant surfaces.
- Any user-stated constraints, such as no network calls or no destructive tests.
Use `rg`, `git diff`, `git log`, LSP, and existing tests before assigning work.
### Phase 1: Independent Hunter Pass
Send one prompt to the 3 hunters:
```text
Audit target:
{target summary}
Context:
{diff, file list, security-sensitive paths, known constraints}
Task:
Find candidate vulnerabilities in your assigned role. For each candidate include:
- title
- affected file/function
- attacker capability
- attack path
- impact
- CWE candidate
- exact evidence
- safe verification idea
Reject generic hardening advice. Return only candidates with a plausible path.
```
Wait for all hunters.
### Phase 2: PoC Pass
Deduplicate hunter candidates. Send the strongest candidates to both PoC engineers.
Each PoC engineer must return:
- Reproduced, falsified, or unsafe-to-run.
- Exact commands, fixtures, or static proof.
- Observed output or reason it fails.
- Severity recommendation using exploitability and impact.
- Downgrade rationale for anything not reproduced.
### Phase 3: Cross-Check
Send the PoC results back to all 5 members.
Ask every member:
- Which findings survive?
- Which findings should be downgraded or removed?
- What remediation is smallest and specific?
- What regression test would prevent recurrence?
### Phase 4: Final Report
Produce this report:
```markdown
## Security Research Result
### Verdict
PASS | PASS WITH FINDINGS | BLOCK
### Scope
- Target:
- Base/diff:
- Commands run:
### Findings
| Severity | Title | CWE | Exploitability | Impact | PoC | Fix |
|----------|-------|-----|----------------|--------|-----|-----|
### Finding Details
For each finding:
- Evidence:
- Attack path:
- PoC:
- Severity rationale:
- Minimal fix:
- Regression check:
### Downgraded or Rejected Candidates
| Candidate | Reason |
|-----------|--------|
### Residual Risk
- What was not tested and why.
```
## Output Rules
- Lead with the verdict.
- Do not bury blocking issues.
- Do not report speculative findings as vulnerabilities.
- Do not claim CVSS precision unless you actually scored the metrics.
- Include exact file paths and commands for every surviving finding.
- If no findings survive PoC, say that plainly and list residual risk.
src/features/builtin-skills/skills.test.ts
+59 -42
View File
@@ -13,8 +13,8 @@ describe("createBuiltinSkills", () => {
// then
const browserSkill = skills.find((s) => s.name === "playwright")
expect(browserSkill).toBeDefined()
expect(browserSkill!.description).toContain("browser")
expect(browserSkill!.mcpConfig?.playwright).toBeDefined()
expect(browserSkill?.description).toContain("browser")
expect(browserSkill?.mcpConfig?.playwright).toBeDefined()
})
test("returns playwright skill when browserProvider is 'playwright'", () => {
@@ -46,7 +46,7 @@ describe("createBuiltinSkills", () => {
const playwrightSkill = skills.find((skill) => skill.name === "playwright")
const agentBrowserSkill = skills.find((skill) => skill.name === "agent-browser")
expect(devBrowserSkill).toBeDefined()
expect(devBrowserSkill!.description).toContain("Browser automation")
expect(devBrowserSkill?.description).toContain("Browser automation")
expect(playwrightSkill).toBeUndefined()
expect(agentBrowserSkill).toBeUndefined()
expect(skillNames).not.toContain("playwright-cli")
@@ -64,9 +64,9 @@ describe("createBuiltinSkills", () => {
const agentBrowserSkill = skills.find((s) => s.name === "agent-browser")
const playwrightSkill = skills.find((s) => s.name === "playwright")
expect(agentBrowserSkill).toBeDefined()
expect(agentBrowserSkill!.description).toContain("browser")
expect(agentBrowserSkill!.allowedTools).toContain("Bash(agent-browser:*)")
expect(agentBrowserSkill!.template).toContain("agent-browser")
expect(agentBrowserSkill?.description).toContain("browser")
expect(agentBrowserSkill?.allowedTools).toContain("Bash(agent-browser:*)")
expect(agentBrowserSkill?.template).toContain("agent-browser")
expect(playwrightSkill).toBeUndefined()
})
@@ -79,13 +79,13 @@ describe("createBuiltinSkills", () => {
const agentBrowserSkill = skills.find((s) => s.name === "agent-browser")
// then - template should contain substantial content (inlined, not fallback)
expect(agentBrowserSkill!.template).toContain("## Quick start")
expect(agentBrowserSkill!.template).toContain("## Commands")
expect(agentBrowserSkill!.template).toContain("agent-browser open")
expect(agentBrowserSkill!.template).toContain("agent-browser snapshot")
expect(agentBrowserSkill?.template).toContain("## Quick start")
expect(agentBrowserSkill?.template).toContain("## Commands")
expect(agentBrowserSkill?.template).toContain("agent-browser open")
expect(agentBrowserSkill?.template).toContain("agent-browser snapshot")
})
test("always includes frontend-ui-ux, git-master, review-work, ai-slop-remover, and security-review skills", () => {
test("always includes frontend-ui-ux, git-master, review-work, ai-slop-remover, security-research, and security-review skills", () => {
// given - both provider options
// when
@@ -99,11 +99,12 @@ describe("createBuiltinSkills", () => {
expect(skills.find((s) => s.name === "git-master")).toBeDefined()
expect(skills.find((s) => s.name === "review-work")).toBeDefined()
expect(skills.find((s) => s.name === "ai-slop-remover")).toBeDefined()
expect(skills.find((s) => s.name === "security-research")).toBeDefined()
expect(skills.find((s) => s.name === "security-review")).toBeDefined()
}
})
test("returns exactly 6 skills regardless of provider", () => {
test("returns exactly 7 skills regardless of provider", () => {
// given
// when
@@ -112,9 +113,9 @@ describe("createBuiltinSkills", () => {
const devBrowserSkills = createBuiltinSkills({ browserProvider: "dev-browser" })
// then
expect(defaultSkills).toHaveLength(6)
expect(agentBrowserSkills).toHaveLength(6)
expect(devBrowserSkills).toHaveLength(6)
expect(defaultSkills).toHaveLength(7)
expect(agentBrowserSkills).toHaveLength(7)
expect(devBrowserSkills).toHaveLength(7)
})
test("should exclude playwright when it is in disabledSkills", () => {
@@ -131,8 +132,9 @@ describe("createBuiltinSkills", () => {
expect(skills.map((s) => s.name)).not.toContain("dev-browser")
expect(skills.map((s) => s.name)).toContain("review-work")
expect(skills.map((s) => s.name)).toContain("ai-slop-remover")
expect(skills.map((s) => s.name)).toContain("security-research")
expect(skills.map((s) => s.name)).toContain("security-review")
expect(skills.length).toBe(5)
expect(skills.length).toBe(6)
})
test("should exclude multiple skills when they are in disabledSkills", () => {
@@ -149,13 +151,14 @@ describe("createBuiltinSkills", () => {
expect(skills.map((s) => s.name)).not.toContain("dev-browser")
expect(skills.map((s) => s.name)).toContain("review-work")
expect(skills.map((s) => s.name)).toContain("ai-slop-remover")
expect(skills.map((s) => s.name)).toContain("security-research")
expect(skills.map((s) => s.name)).toContain("security-review")
expect(skills.length).toBe(4)
expect(skills.length).toBe(5)
})
test("should return an empty array when all skills are disabled", () => {
// #given
const options = { disabledSkills: new Set(["playwright", "frontend-ui-ux", "git-master", "review-work", "ai-slop-remover", "security-review"]) }
const options = { disabledSkills: new Set(["playwright", "frontend-ui-ux", "git-master", "review-work", "ai-slop-remover", "security-research", "security-review"]) }
// #when
const skills = createBuiltinSkills(options)
@@ -172,7 +175,7 @@ describe("createBuiltinSkills", () => {
const skills = createBuiltinSkills(options)
// #then
expect(skills.length).toBe(6)
expect(skills.length).toBe(7)
})
test("review-work skill has correct structure", () => {
@@ -184,13 +187,13 @@ describe("createBuiltinSkills", () => {
// #then
expect(reviewWork).toBeDefined()
expect(reviewWork!.description).toContain("review")
expect(reviewWork!.template).toContain("5-Agent Parallel Review Orchestrator")
expect(reviewWork!.template).toContain("Goal & Constraint Verification")
expect(reviewWork!.template).toContain("QA")
expect(reviewWork!.template).toContain("Code Quality")
expect(reviewWork!.template).toContain("Security")
expect(reviewWork!.template).toContain("Context Mining")
expect(reviewWork?.description).toContain("review")
expect(reviewWork?.template).toContain("5-Agent Parallel Review Orchestrator")
expect(reviewWork?.template).toContain("Goal & Constraint Verification")
expect(reviewWork?.template).toContain("QA")
expect(reviewWork?.template).toContain("Code Quality")
expect(reviewWork?.template).toContain("Security")
expect(reviewWork?.template).toContain("Context Mining")
})
test("review-work skill explains Codex tool compatibility before OpenCode orchestration examples", () => {
@@ -199,8 +202,9 @@ describe("createBuiltinSkills", () => {
// #when
const reviewWork = skills.find((s) => s.name === "review-work")
const compatibilityIndex = reviewWork!.template.indexOf("## Codex Harness Tool Compatibility")
const opencodeExampleIndex = reviewWork!.template.search(/\b(?:background_output|team_[a-z_]+|task)\s*\(/)
const reviewWorkTemplate = reviewWork?.template ?? ""
const compatibilityIndex = reviewWorkTemplate.indexOf("## Codex Harness Tool Compatibility")
const opencodeExampleIndex = reviewWorkTemplate.search(/\b(?:background_output|team_[a-z_]+|task)\s*\(/)
// #then
expect(compatibilityIndex >= 0).toBe(true)
@@ -216,23 +220,36 @@ describe("createBuiltinSkills", () => {
// #then
expect(aiSlopRemover).toBeDefined()
expect(aiSlopRemover!.description).toContain("AI-generated code smells")
expect(aiSlopRemover!.template).toContain("DETECTION CRITERIA")
expect(aiSlopRemover!.template).toContain("SAFETY RULES")
expect(aiSlopRemover?.description).toContain("AI-generated code smells")
expect(aiSlopRemover?.template).toContain("DETECTION CRITERIA")
expect(aiSlopRemover?.template).toContain("SAFETY RULES")
})
test("security-review skill has correct structure", () => {
test("security-research skill has correct structure", () => {
// #given - default options
// #when
const skills = createBuiltinSkills()
const securityResearch = skills.find((skill) => skill.name === "security-research")
// #then
expect(securityResearch?.description).toContain("security research")
expect(securityResearch?.template).toContain("Security Research - Team Mode Vulnerability Audit")
expect(securityResearch?.template).toContain('name: "security-research"')
expect(securityResearch?.template).toContain("Security Research Result")
})
test("security-review skill remains a runtime alias for security-research", () => {
// #given - default options
// #when
const skills = createBuiltinSkills()
const securityReview = skills.find((skill) => skill.name === "security-review")
const securityResearch = skills.find((skill) => skill.name === "security-research")
// #then
expect(securityReview?.description).toContain("security review")
expect(securityReview?.template).toContain("Team Mode Vulnerability Audit")
expect(securityReview?.template).toContain("team_create")
expect(securityReview?.template).toContain("Security Review Result")
expect(securityReview?.description).toContain("Alias for security-research")
expect(securityReview?.template).toBe(securityResearch?.template)
})
test("returns playwright-cli skill when browserProvider is 'playwright-cli'", () => {
@@ -246,9 +263,9 @@ describe("createBuiltinSkills", () => {
const playwrightSkill = skills.find((s) => s.name === "playwright")
const agentBrowserSkill = skills.find((s) => s.name === "agent-browser")
expect(playwrightSkill).toBeDefined()
expect(playwrightSkill!.description).toContain("browser")
expect(playwrightSkill!.allowedTools).toContain("Bash(playwright-cli:*)")
expect(playwrightSkill!.mcpConfig).toBeUndefined()
expect(playwrightSkill?.description).toContain("browser")
expect(playwrightSkill?.allowedTools).toContain("Bash(playwright-cli:*)")
expect(playwrightSkill?.mcpConfig).toBeUndefined()
expect(agentBrowserSkill).toBeUndefined()
})
@@ -261,8 +278,8 @@ describe("createBuiltinSkills", () => {
const skill = skills.find((s) => s.name === "playwright")
// then
expect(skill!.template).toContain("playwright-cli open")
expect(skill!.template).toContain("playwright-cli snapshot")
expect(skill!.template).toContain("playwright-cli click")
expect(skill?.template).toContain("playwright-cli open")
expect(skill?.template).toContain("playwright-cli snapshot")
expect(skill?.template).toContain("playwright-cli click")
})
})
src/features/builtin-skills/skills.ts
+2
View File
@@ -10,6 +10,7 @@ import {
devBrowserSkill,
reviewWorkSkill,
aiSlopRemoverSkill,
securityResearchSkill,
securityReviewSkill,
teamModeSkill,
} from "./skills/index"
@@ -40,6 +41,7 @@ export function createBuiltinSkills(options: CreateBuiltinSkillsOptions = {}): B
gitMasterSkill,
reviewWorkSkill,
aiSlopRemoverSkill,
securityResearchSkill,
securityReviewSkill,
]
src/features/builtin-skills/skills/index.ts
+1
View File
@@ -5,5 +5,6 @@ export { gitMasterSkill } from "./git-master"
export { devBrowserSkill } from "./dev-browser"
export { reviewWorkSkill } from "./review-work"
export { aiSlopRemoverSkill } from "./ai-slop-remover"
export { securityResearchSkill } from "./security-research"
export { securityReviewSkill } from "./security-review"
export * from "./team-mode"
src/features/builtin-skills/skills/security-research.ts
+9
View File
@@ -0,0 +1,9 @@
import type { BuiltinSkill } from "../types"
import securityResearchTemplate from "../security-research/SKILL.md" with { type: "text" }
export const securityResearchSkill: BuiltinSkill = {
name: "security-research",
description:
"Team Mode security research skill. Orchestrates 3 vulnerability hunters and 2 PoC engineers to audit a codebase in parallel, prove exploitability, classify root causes, and calibrate severity by actual exploitability. Use for security review, vulnerability research, exploitability audit, pre-release security check, threat model validation, and `/security-research`. Triggers: 'security-research', 'security research', 'security review', 'vulnerability audit', 'exploitability audit', '보안 리뷰', '취약점 감사'.",
template: securityResearchTemplate,
}
src/features/builtin-skills/skills/security-review.ts
+3 -201
View File
@@ -1,206 +1,8 @@
import type { BuiltinSkill } from "../types"
import { securityResearchSkill } from "./security-research"
export const securityReviewSkill: BuiltinSkill = {
name: "security-review",
description:
"Team Mode security review skill. Orchestrates 3 vulnerability hunters and 2 PoC engineers to audit a codebase in parallel, prove exploitability, classify root causes, and calibrate severity by actual exploitability. Use for security review, vulnerability research, exploitability audit, pre-release security check, threat model validation, and `/security-review`. Triggers: 'security-review', 'security review', 'security research', 'vulnerability audit', 'exploitability audit', '보안 리뷰', '취약점 감사'.",
template: `# Security Review - Team Mode Vulnerability Audit
Use this skill to run a parallel security audit that separates real exploitability from generic concern. The team has 3 vulnerability hunters and 2 PoC engineers.
## Hard Preconditions
Before starting, verify:
1. \`team_*\` tools are available. If not, stop and tell the user:
\`security-review requires team-mode. Set team_mode.enabled: true in your oh-my-openagent config, restart opencode, then retry.\`
2. You are in the main session, not a background subagent.
3. You have a concrete target: repository, diff range, PR, release candidate, path list, or threat surface.
If the user provided no target, audit the current repository and current branch diff against its upstream or merge base. If there is no diff, audit the security-sensitive surfaces in the working tree.
## Severity Standard
Use these references as the scoring frame:
- CWE for root-cause weakness classification: https://cwe.mitre.org/
- OWASP WSTG for test methodology: https://devguide.owasp.org/en/06-verification/01-guides/01-wstg/
- OWASP ASVS for control verification: https://owasp.org/www-project-application-security-verification-standard/
- CVSS v4.0 for exploitability and impact scoring: https://www.first.org/cvss/v4.0/specification-document
Rules:
- No severity without an attack path.
- No critical or high finding without concrete exploit preconditions and impact.
- Keep CWE category separate from severity.
- Prefer a small, reproducible PoC over theoretical language.
- Never run destructive exploits against real services or third-party systems.
- Use local fixtures, toy payloads, dry runs, or static proof when real execution would be unsafe.
## Team Roster
Create one Team Mode run with these 5 members:
| Member | Kind | Category | Role |
|--------|------|----------|------|
| \`surface-hunter\` | category | \`deep\` | Map entry points, trust boundaries, and reachable attack surfaces. |
| \`auth-data-hunter\` | category | \`ultrabrain\` | Hunt auth, authorization, data isolation, injection, and secret handling flaws. |
| \`runtime-supply-hunter\` | category | \`unspecified-high\` | Hunt filesystem, subprocess, archive, dependency, hook, MCP, and config risks. |
| \`poc-engineer-a\` | category | \`unspecified-high\` | Build minimal PoCs for the strongest candidate findings. |
| \`poc-engineer-b\` | category | \`deep\` | Independently reproduce, falsify, or downgrade candidate findings. |
Call \`team_create\` with an inline spec:
\`\`\`typescript
team_create({
inline_spec: {
name: "security-review",
description: "Parallel exploitability-driven security review team.",
members: [
{
name: "surface-hunter",
kind: "category",
category: "deep",
prompt: "You map attack surface. Enumerate entry points, trust boundaries, attacker-controlled inputs, data sinks, privilege transitions, and sensitive assets. Return evidence with file paths and exact functions. Do not assign severity unless you can name an attack path."
},
{
name: "auth-data-hunter",
kind: "category",
category: "ultrabrain",
prompt: "You hunt auth, authorization, tenant/data isolation, injection, SSRF, credential exposure, and confused-deputy flaws. Reason from attacker capability to impact. Return only findings with concrete exploit preconditions, CWE candidates, and verification steps."
},
{
name: "runtime-supply-hunter",
kind: "category",
category: "unspecified-high",
prompt: "You hunt filesystem, subprocess, archive extraction, dependency, hook execution, MCP, config, and environment-variable risks. Check path traversal, command injection, unsafe downloads, permission boundaries, and supply-chain assumptions. Cite file paths and commands used."
},
{
name: "poc-engineer-a",
kind: "category",
category: "unspecified-high",
prompt: "You build minimal safe PoCs for candidate findings. Use toy inputs and local-only execution. Your job is to prove or disprove exploitability, not to broaden scope. Report exact reproduction steps and expected output."
},
{
name: "poc-engineer-b",
kind: "category",
category: "deep",
prompt: "You independently reproduce candidate findings and try to falsify them. Downgrade anything without a working path. If a PoC is unsafe to run, design a safe static or dry-run proof and explain the limit."
}
]
}
})
\`\`\`
If a category is unavailable, retry once by replacing only that category with \`unspecified-high\`. Do not reduce the team below 5 members.
## Workflow
### Phase 0: Scope and Baseline
Collect:
- Target scope and reason for audit.
- Branch, base ref, diff, and changed files if this is a change review.
- Security-sensitive directories and files if this is a full-repo audit.
- Existing tests and commands that exercise relevant surfaces.
- Any user-stated constraints, such as no network calls or no destructive tests.
Use \`rg\`, \`git diff\`, \`git log\`, LSP, and existing tests before assigning work.
### Phase 1: Independent Hunter Pass
Send one prompt to the 3 hunters:
\`\`\`text
Audit target:
{target summary}
Context:
{diff, file list, security-sensitive paths, known constraints}
Task:
Find candidate vulnerabilities in your assigned role. For each candidate include:
- title
- affected file/function
- attacker capability
- attack path
- impact
- CWE candidate
- exact evidence
- safe verification idea
Reject generic hardening advice. Return only candidates with a plausible path.
\`\`\`
Wait for all hunters.
### Phase 2: PoC Pass
Deduplicate hunter candidates. Send the strongest candidates to both PoC engineers.
Each PoC engineer must return:
- Reproduced, falsified, or unsafe-to-run.
- Exact commands, fixtures, or static proof.
- Observed output or reason it fails.
- Severity recommendation using exploitability and impact.
- Downgrade rationale for anything not reproduced.
### Phase 3: Cross-Check
Send the PoC results back to all 5 members.
Ask every member:
- Which findings survive?
- Which findings should be downgraded or removed?
- What remediation is smallest and specific?
- What regression test would prevent recurrence?
### Phase 4: Final Report
Produce this report:
\`\`\`markdown
## Security Review Result
### Verdict
PASS | PASS WITH FINDINGS | BLOCK
### Scope
- Target:
- Base/diff:
- Commands run:
### Findings
| Severity | Title | CWE | Exploitability | Impact | PoC | Fix |
|----------|-------|-----|----------------|--------|-----|-----|
### Finding Details
For each finding:
- Evidence:
- Attack path:
- PoC:
- Severity rationale:
- Minimal fix:
- Regression check:
### Downgraded or Rejected Candidates
| Candidate | Reason |
|-----------|--------|
### Residual Risk
- What was not tested and why.
\`\`\`
## Output Rules
- Lead with the verdict.
- Do not bury blocking issues.
- Do not report speculative findings as vulnerabilities.
- Do not claim CVSS precision unless you actually scored the metrics.
- Include exact file paths and commands for every surviving finding.
- If no findings survive PoC, say that plainly and list residual risk.
`,
description: `Alias for security-research. ${securityResearchSkill.description}`,
template: securityResearchSkill.template,
}