agentleFS
Sign inStart free
IgorGanapolskyIgorGanapolsky/ThumbGate/.agents/skills/salt-code-policy-honesty-not-clone/SKILL.md

salt-code-policy-honesty-not-clone

Skill29 starsChanged 7 months agoMITView on GitHub

Salt Code (getsaltcode.com by Salt Security) provides 40 security policies as prompt-time context injection, not an execution firewall. Steal the 40-policy taxonomy, OpenAPI query-string auth diode, and MCP credential hygiene onto ThumbGate PreToolUse rails. Never install Salt Code MCP or route traffic to mcp.getsaltcode.com. Slash: /salt-code-policy-honesty-not-clone.

Create a free account to mount
Files
SKILL.md

Salt Code policy honesty — compare, do not clone

Goal

Compare Salt Code's advisory prompt-time context injection against ThumbGate's deterministic execution-layer PreToolUse firewall. Steal the 40-policy taxonomy (OWASP API Top 10, OWASP LLM Top 10, MCP Security, OpenAPI Hygiene) and enforce fail-closed checks when security rules exist only in context without PreToolUse enforcement.

Constraints

NEVERALWAYS
Route MCP traffic to mcp.getsaltcode.comRun 100% local stdio/subshell inspection
Rely on prompt-only rules (.cursorrules) to stop attacksEnforce deterministic PreToolUse execution gates
Pass API keys or secrets in URL query parameters (?apiKey=)Enforce Authorization: Bearer headers
Treat Salt Code as a drop-in substitute for ThumbGateDistinguish advisory context from active execution firewall

HARD fail closed. REFUSE SKU clones and cloud dependencies.

Reference

  • https://getsaltcode.com/ (Salt Security, 2026)
  • scripts/salt-code-policy-honesty.js
  • docs/agents/salt-code-policy-honesty.md
  • ThumbGate Completion Claim Contract in AGENTS.md / GEMINI.md

Examples (show, don't tell)

Weak: "Salt Code protects agents by giving them a system prompt with 40 rules."

Gold:

$ npx thumbgate salt-code-policy-honesty --inspect-context=.cursorrules --json
# Flags context_injection_without_pretool_enforcement if PreToolUse is unconfigured.

$ npx thumbgate salt-code-policy-honesty --map-only --json
# Dumps complete 40-policy taxonomy mapped to deterministic PreToolUse gates.

$ npx thumbgate salt-code-policy-honesty --eval-prompt="Design me a delete user API with token in query string" --json
# Intercepts prompt violations and emits proactive, safe alternative designs.

$ npx thumbgate salt-code-policy-honesty --onboarding=cursor --deeplinks --vibe-stats --json
# Emits 1-click cursor:// and vscode: deep links plus empirical vibe-coding CVE stats.

Procedures

# Print taxonomy and rail map
npx thumbgate salt-code-policy-honesty --map-only --json

# Intercept and rewrite insecure prompts before code generation
npx thumbgate salt-code-policy-honesty --eval-prompt="Design delete user endpoint with token in query" --json

# 1-Click onboarding deep links and configs for 16 AI coding agents
npx thumbgate salt-code-policy-honesty --onboarding=cursor --deeplinks --vibe-stats --json

# Audit code, OpenAPI spec, and MCP config
npx thumbgate salt-code-policy-honesty \
  --inspect-code=src/api.js \
  --inspect-api=openapi/spec.yaml \
  --inspect-mcp=.mcp.json \
  --strict

# Run regression test suite
node --test tests/salt-code-policy-honesty.test.js

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.

Check your setup, free