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.
git clone --depth 1 https://github.com/iamgoatedaf/lighter-mcpWrote 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/commands/iamgoatedaf/lighter-mcp/lighter-audit)<a href="https://agentmods.dev/commands/iamgoatedaf/lighter-mcp/lighter-audit"><img src="https://agentmods.dev/badge/commands/iamgoatedaf/lighter-mcp/lighter-audit/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/commands/iamgoatedaf/lighter-mcp/lighter-audit"><img src="https://agentmods.dev/badge/commands/iamgoatedaf/lighter-mcp/lighter-audit.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00027 | $0.00584 |
| Opus 5 | $0.00014 | $0.00292 |
| Sonnet 5 | $0.00005 | $0.00117 |
| Haiku 4.5 | $0.00003 | $0.00058 |
Grade A, and why
lighter-audit 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 10d 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.
What it actually says
/lighter-audit
The audit log lives at ~/.lighter/lighter-mcp/audit.jsonl (append-only
JSONL, one record per tool call, secrets pre-redacted by the server).
Default behavior: show the last 20 records of the last 24 hours.
Optional arguments the user might pass in plain language:
- "last N hours" / "since N hours ago"
- "only failures"
- "tool lighter_live_market_order" (or any prefix like
lighter_live_*) - "mode live" / "mode paper"
Steps:
-
Parse the user's natural-language filters. Defaults if absent:
hours=24,limit=20, no tool/mode filter, all results. -
Read the log with a small shell snippet (do not load the whole file into context — it can be large):
python3 - <<'PY' import json, os, time, pathlib HOURS = 24 # replace with parsed value LIMIT = 20 # replace with parsed value TOOL = None # e.g. "lighter_live_market_order" or prefix MODE = None # "live" / "paper" / etc. OK = None # True for only-success, False for only-failures, None for both cutoff = time.time() - HOURS * 3600 p = pathlib.Path(os.path.expanduser("~/.lighter/lighter-mcp/audit.jsonl")) if not p.is_file(): print("[]"); raise SystemExit rows = [] with p.open() as fh: for line in fh: try: r = json.loads(line) except Exception: continue if r.get("ts", 0) < cutoff: continue if TOOL and not r.get("tool", "").startswith(TOOL): continue if MODE and r.get("mode") != MODE: continue if OK is not None and bool(r.get("ok")) != OK: continue rows.append(r) rows = rows[-LIMIT:] print(json.dumps(rows, indent=2)) PY -
Render a compact table: timestamp (local time), tool, mode, ok/error, one-line summary of args (symbol/amount where present).
-
If a row has
error, show theerrorandcategoryverbatim — these are the safety/confirmation envelopes and they are exactly what the user wants to inspect.
Never modify or rotate the audit file from this command.
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.
- 10d ago First seen · 60 lines · 27 tokens per session scan A 1af9f3fcceff
lighter-audit is a command published in the GitHub repository iamgoatedaf/lighter-mcp (1 stars, last pushed 4mo ago), licensed MIT. It adds 27 tokens to every session and 584 once invoked, about $0.0001 per session on Opus 5. 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 commands, from other repositories
audit-strict
Multi-pass consensus audit — runs the audit twice with different prompts, only reports consensus findings. Aggressively cuts false positives.
explain
Explain a finding in plain English, with examples. Optionally tailored to a beginner audience.
data-mesh-contract
Create federated data product contracts for mesh architectures with SLAs, governance, and interoperability guarantees (project).
audit-web
A command for requesting a website security assessment through an MCP tool, a way for an agent to call an external service. It requires proof of permission before any network checks.
atrs
Generate Algorithmic Transparency Recording Standard (ATRS) record for AI/algorithmic tools.
fr-rgpd
You are helping an enterprise architect generate a French CNIL Compliance Assessment — the French-specific GDPR layer applied by the CNIL (Commission Nationale de l'Informatique et des Libertés). Run this after /arckit-eu:eu-rgpd to add French obligations that go beyond the EU GDPR baseline.