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 skills/sonomoshq/canary/tokennpx skills add sonomoshq/Canary --skill tokengit clone --depth 1 https://github.com/sonomoshq/CanaryWrote 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/skills/sonomoshq/canary/token)<a href="https://agentmods.dev/skills/sonomoshq/canary/token"><img src="https://agentmods.dev/badge/skills/sonomoshq/canary/token.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 | $0.00035 | $0.01270 |
| Opus 5 | $0.00017 | $0.00635 |
| Sonnet 5 | $0.00007 | $0.00254 |
| Haiku 4.5 | $0.00003 | $0.00127 |
Grade A, and why
token 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 4d 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 — 117 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Canary Tokens — Active Defense
Every other part of Canary is passive: 38 regex detectors guess whether
something looks like PII or a secret, from its shape. Canary Tokens flip
that around. Canary mints a fake-but-realistic secret itself, plants it,
and remembers the exact value it planted. When that exact value later
shows up in something Claude reads, it isn't a guess about shape — it's
a literal string match (grep -F) against a value nobody else could have
produced. 38 detectors guess. This one knows.
A "trip" is recorded at confidence: certain, distinct from every other
entry in the leak log, which is at best high/medium.
Instructions
Parse $ARGUMENTS for a subcommand and dispatch to the canary-token
CLI (bundled at ${CLAUDE_PLUGIN_ROOT}/bin/canary-token, already on
PATH as canary-token). Prefer running it directly; fall back to
bash "${CLAUDE_PLUGIN_ROOT}/bin/canary-token" ... if the bare command
isn't found.
new <aws|card|ssn|env|dburl|freeform> [label]
Mint a fresh fake decoy and register it as an armed canary — nothing is written to disk yet.
canary-token new env "staging API key"
For freeform, pass the decoy string itself in place of a label:
canary-token new freeform "Project Nightjar"
Types and what each mimics:
aws— anAKIA...access key + a paired secret key (also matches Canary's ownaws_access_key/aws_secret_keydetectors)card— a Luhn-valid 16-digit card number (also matchescredit_card)ssn— an SSN in the 900-999 area, which the SSA has never issued to a real person, so it can never collide with someone's real SSNenv— a high-entropy API-key-shaped string (also matchesgeneric_secretwhen planted with a keyword likeAPI_KEY=)dburl— a decoypostgres://user:pass@host/dbconnection string (also matchesdb_url_credentials)freeform— any arbitrary string (a fake codename, trade secret, etc.) — detected only by Canary's certain match, no regex layer involved
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.
- 4d ago First seen · 117 lines · 35 tokens per session scan A 837ee31ed6ec
token is a skill published in the GitHub repository sonomoshq/Canary (2 stars, last pushed 1mo ago), licensed MIT. It adds 35 tokens to every session and 1,270 once invoked, about $0.0002 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 skills, from other repositories
compare
Compare multiple alternatives using explicit criteria, weighted scoring, and tradeoff analysis. Use when choosing between options, evaluating alternatives, or making decisions.
critique
Find failure modes, edge cases, ambiguities, and exploit paths in plans, code, or designs. Use when reviewing proposals, auditing security, stress-testing logic, or validating assumptions.
explain
Produce clear reasoning with assumptions, causal chains, and evidence. Use when clarifying decisions, teaching concepts, justifying recommendations, or documenting rationale.
integrate
Combine heterogeneous data sources into a unified model with conflict resolution, schema alignment, and provenance tracking. Use when merging data from multiple systems, consolidating information, or building comprehensive views.
audit
Produce a comprehensive audit trail of actions, tools used, changes made, and decision rationale. Use when recording compliance evidence, tracking changes, or documenting decision lineage.
gap-analysis-workflow
Identify capability gaps and propose new skills with prioritization. Use when analyzing missing capabilities, planning skill development, performing ontology expansion, or assessing coverage.