Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add agents/darkiceinteractive/mcp-conductor/j-x4-tokenizationgit clone --depth 1 https://github.com/darkiceinteractive/mcp-conductorWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/darkiceinteractive/mcp-conductor/j-x4-tokenization)<a href="https://agentmods.dev/agents/darkiceinteractive/mcp-conductor/j-x4-tokenization"><img src="https://agentmods.dev/badge/agents/darkiceinteractive/mcp-conductor/j-x4-tokenization.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00000 | $0.00805 |
| Opus 5 | $0.00000 | $0.00402 |
| Sonnet 5 | $0.00000 | $0.00161 |
| Haiku 4.5 | $0.00000 | $0.00081 |
Grade A, and why
J-x4-tokenization scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 59 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Agent J — Workstream X4 PII tokenization
You add per-server redact.response config that strips sensitive values from upstream responses before they enter the sandbox or Claude's context. Reverse map sandbox-local; mcp.detokenize(value) for outbound calls.
Setup
pwd # /mcp-executor-darkice-worktrees/J-x4
git branch --show-current # feature/v3-x4-tokenization
git fetch origin
git rebase origin/feature/v3-phase-0-1 # registry needed
git rebase origin/feature/v3-x1-passthrough # X1 routing annotations side-by-side with redact
npm install && npm run test:run
Read /Users/mattcrombie/.claude/plans/read-this-analysis-from-rosy-whale.md §3 Part C "Workstream X4".
Append start checkpoint to STATUS.md.
Scope
Built-in matchers ONLY this cut (decision recorded — no inline-regex this sprint):
- email (RFC 5322 simplified)
- phone (loose international, e.g.
+CC NNNN NNN NNN,(NNN) NNN-NNNN) - SSN (
NNN-NN-NNNNand 9-digit run with context) - credit card (Luhn-validated, 13–19 digits with optional separators)
- IBAN (basic checksum)
- IPv4 + IPv6
Files:
src/utils/tokenize.ts(NEW) — pure tokenizer; reuses Phase 1.6 redact infrastructure (src/utils/redact.tsalready exists for log redaction) but generalises it to data-flow with a reverse map.src/hub/mcp-hub.ts— apply tokenizer to result before return whenToolDefinition.redact.responseannotation present.src/runtime/executor.ts— sandbox preamble:mcp.detokenize(value)looks up the reverse map. Reverse map is per-execution (not shared acrossexecute_codecalls) so a token can't survive past the call that minted it.
Annotation read from ToolDefinition.redact (set via update_server MCP tool [Agent I's territory] or directly in ~/.mcp-conductor.json).
Acceptance
- Server returning
{"email":"[email protected]", "phone":"+61 412 345 678"}surfaces{"email":"[EMAIL_1]", "phone":"[PHONE_1]"}in the sandbox. mcp.detokenize("[EMAIL_1]")returns[email protected]inside the sameexecute_codecall.- Token survives a within-call outbound MCP call (e.g.
mcp.server('crm').call('lookup', {email: mcp.detokenize('[EMAIL_1]')})works). - Token does NOT detokenize when returned as the final result (Claude sees
[EMAIL_1]). - A subsequent
execute_codecall cannot detokenize tokens from a prior call. test/unit/tokenize.test.ts,test/integration/tokenize-flow.test.ts(~12 cases) pass.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 59 lines · 0 tokens per session scan A 0619ae5621c8
J-x4-tokenization is an agent published in the GitHub repository darkiceinteractive/mcp-conductor (0 stars, last pushed 2mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 805 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other agents, from other repositories
ci-local
Run all CI checks locally and systematically fix any failures. Replicates GitHub workflow validations.
work-verifier
Validates completed work. Use after tasks are marked done to confirm implementations are functional.
Demonstrate
Agent for demonstrating VS Code features.
playwright-test-generator
Use this agent when you need to create automated browser tests using Playwright Examples: Context: User wants to generate a test for the test plan item.
analyzer
Analyze blind comparison results to understand WHY the winner won and generate improvement suggestions.
grader
Evaluate expectations against an execution transcript and outputs.