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 skills add corveil/crow --skill verifygit clone --depth 1 https://github.com/corveil/crowWrote 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/corveil/crow/verify)<a href="https://agentmods.dev/skills/corveil/crow/verify"><img src="https://agentmods.dev/badge/skills/corveil/crow/verify/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/skills/corveil/crow/verify"><img src="https://agentmods.dev/badge/skills/corveil/crow/verify.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Data Exfiltration · line 40 Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.Fix: Verify the destination URL is trusted and necessary. Remove or replace with documented APIs. Ensure no secrets, tokens, or PII are transmitted.
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 | $0.00030 | $0.00854 |
| Opus 5.5 | $0.00012 | $0.00342 |
| Sonnet 5.5 | $0.00006 | $0.00171 |
| Haiku 4.5 | $0.00003 | $0.00085 |
Grade A, and why
verify scanned grade A with 1 finding 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 1mo 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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
description: Drive CrowTelemetry's OTLP ingest end-to-end — boot the real receiver, POST OTLP JSON with curl, inspect the SQLite db. What it actually says
Verifying CrowTelemetry changes
The package's runtime surface is TelemetryService (OTLPReceiver HTTP
listener on localhost + TelemetryDatabase SQLite file). There is no
standalone binary — build a throwaway harness executable that depends on this
package and boots the service:
// Sources/otlp-harness/main.swift — usage: otlp-harness <port> <data-dir>
import Foundation
import CrowTelemetry
let service = try TelemetryService(port: UInt16(CommandLine.arguments[1])!,
dataDirectory: CommandLine.arguments[2]) { id in
print("[harness] data received for session \(id)")
}
Task { try await service.start() }
dispatchMain()
Harness Package.swift: swift-tools-version 6.0, platforms: [.macOS(.v14)],
.package(path: "<repo>/Packages/CrowTelemetry"). Build with swift build,
run in the background, then drive it.
Gotchas:
-
Ingest is OTLP HTTP/JSON only (
OTEL_EXPORTER_OTLP_PROTOCOL=http/json), paths/v1/metricsand/v1/logs, POST only. One request per connection (the receiver closes it after responding).Content-LengthandTransfer-Encoding: chunkedboth work; a compressed body or a non-JSONContent-Typeis rejected by name. -
Content-Type: application/jsonis required on POST — the receiver fails closed. Pass it explicitly with curl, which otherwise sendsapplication/x-www-form-urlencodedfor-dand gets a 400:curl -X POST http://localhost:<port>/v1/logs \ -H 'Content-Type: application/json' --data-binary @payload.json -
The resource must carry
crow.session.id(a UUID string) inresource.attributes, or the payload is silently skipped. -
Datapoint values:
asDouble(number) orasInt(string or number). -
Sum metrics take
aggregationTemporality(1=delta, 2=cumulative, or theAGGREGATION_TEMPORALITY_*name) andisMonotonic; cumulative sums are normalized to deltas at insert. -
Decoding is deliberately tolerant (#823): scalars are accepted in either their number or string form, and a malformed record is skipped with a
skipped malformed N logRecordslog line rather than failing the export. -
Event names arrive bare in the
event.nameattribute (user_prompt); they are qualified toclaude_code.user_promptat ingest, which is the formcountEventsandturnAnalyticsmatch on. -
Decode failures log the
DecodingErrorcoding path, which names the offending field. SetCROW_TELEMETRY_LOG_RAW_BODIES=1to also log a truncated raw body — off by default because bodies carry account IDs and emails. Repeat failures are throttled to one report per minute per path.
Minimal payload:
{"resourceMetrics":[{"resource":{"attributes":[{"key":"crow.session.id","value":{"stringValue":"<UUID>"}}]},
"scopeMetrics":[{"metrics":[{"name":"claude_code.cost.usage",
"sum":{"aggregationTemporality":2,"isMonotonic":true,"dataPoints":[{"asDouble":1.0}]}}]}]}]}
Observe results with the sqlite3 CLI against <data-dir>/telemetry.db:
SELECT metric_name, value, attributes_json FROM metrics ORDER BY id and
compare SUM(value) per metric to the expected total. Events land in the
events table via /v1/logs.
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.
- 1mo ago First seen · 72 lines · 30 tokens per session scan A 83a5ba7d9949
verify is a skill published in the GitHub repository corveil/crow (20 stars, last pushed today), licensed Apache-2.0. It adds 30 tokens to every session and 854 once invoked, about $0.0001 per session on Opus 5.5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
orca-emulator
This discovery stub loads the version-matched guide from the Orca executable used for this session.
verify-ui-change-in-cloud
Verifies user-facing Warp client changes by spawning a cloud agent with computer use to test Warp. Use only when the user explicitly requested computer-use verification or accepted an offer to run it, and ONLY in non-sandboxed environments and local environments. Triggers a cloud agent that runs the test-warp-ui skill.
test-warp-ui
Guides testing Warp UI features and changes using the computer use tool. Use this skill only when computer-use testing was requested (explicit request or accepted offer) and the computeruse tool is available to the agent. Covers launching Warp and verifying UI behavior.
tui-test
Control, inspect, test, and record real terminal apps with the tui-test CLI or its Rust, Python, and JavaScript APIs. Use for shells, TUI apps, text and style locators, keyboard or mouse input, terminal state, clipboard waits, screenshots, recordings, snapshots, and live terminal sessions.
blueprint-refactor-tranche
Take ONE target from the plan and land it without changing behaviour: guard it on the untouched code, move it verbatim, prove the equivalence by running both, mutation-sweep the tests, pass every gate, open a PR and merge it on green CI — or decline it with the cost written down.
blueprint-cloudflare-security
Review the finished Worker against OWASP Top 10:2025, fix what is exploitable, prove each fix with a test, and report.