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 aahlijia/dekko --skill dekko-verifygit clone --depth 1 https://github.com/aahlijia/dekkoWrote 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/aahlijia/dekko/dekko-verify)<a href="https://agentmods.dev/skills/aahlijia/dekko/dekko-verify"><img src="https://agentmods.dev/badge/skills/aahlijia/dekko/dekko-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/aahlijia/dekko/dekko-verify"><img src="https://agentmods.dev/badge/skills/aahlijia/dekko/dekko-verify.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 | $0.00151 | $0.01485 |
| Opus 5 | $0.00076 | $0.00743 |
| Sonnet 5 | $0.00030 | $0.00297 |
| Haiku 4.5 | $0.00015 | $0.00148 |
Grade A, and why
dekko-verify 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 5d 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 — 101 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Verifying a low-confidence dekko answer
dekko's call-graph resolution is real but conditional — repeated
hands-on evaluation rounds (see this repo's own
test-repos/reports/) keep finding the same failure shape: a
confident wrong answer, not a visible error. A caller trusts "0
callers" more than it should. This skill exists to catch that before
it leads to deleting live code or missing a real impact.
When to double-check before trusting a result
Reach for one targeted grep -rn <name> (not a full re-read) as a
sanity check — not a full re-verification — when any of these apply:
- A cross-package/cross-module qualified call is involved.
pkg.Func()-style calls (Go),namespace::func()(C++), or any call where the receiver is a same-repo package/module rather than a local variable are a known resolver blind spot — confirmed missing 4 real call sites on a live repo as recently as this project's own round-13 eval. Same caution applies to trait/interface dispatch (Rustdyn Traitcalls, Java/Kotlin interface methods) — the resolver ladder only reliably matches an explicitType::method()orType.method()form. - The result doesn't disclose ambiguity. A real ambiguous call
should say so (
N call(s) resolved ambiguously), not just be silently absent from the count. If a symbol you expect to be widely used shows a low count with no ambiguity disclosure, that's more suspicious than a low count with one. - The repo has any unsupported/partially-parsed language files —
check
dekko statsor the map-build summary for an "unsupported" note. Files dekko can't parse are tracked (not silently dropped), but a symbol only ever called from an unparsed file will still read as zero-caller. get_callers/get_calleesused their default--no-testsfilter. An empty result may just mean "no non-test callers" — check whetherinclude_tests/--include-testswas applied before concluding dead code.- A dense-repo common short method name (
new,then,map,iter_mut, or similarly generic names in a 10k+-symbol repo) — resolver precision degrades under high symbol density; treat a count from these as directional, not exact. - You're about to delete or rename based on
dekko unused's dead-code list. Same blind spots apply; a callback passed by reference rather than called directly, or a call from an unparsed file, can both read as "no inbound calls." - A heritage or throws-provenance result labels something
(external).query supertypes/subtypesandquery throwscan mislabel an in-repo type-alias-as-heritage-base or a pattern-bound rethrow as a fake external entry when the extractor doesn't yet model that language's specific syntax shape (confirmed historically on TSimplements <type-alias>and Java 16+instanceof-pattern rethrows — both since fixed, but the general failure shape, a present result that's mislabeled rather than a missing one, can recur in a new syntax shape any of dekko's language extractors hasn't seen yet). An(external)/(unresolved)label on a name you're confident is first-party code is worth aquery symbol <name>check before trusting it.
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.
- 5d ago Changed · +7 lines 3ccc6c95f1c2
- 9d ago First seen · 94 lines · 151 tokens per session scan A 5ca3145b812e
dekko-verify is a skill published in the GitHub repository aahlijia/dekko (3 stars, last pushed 8d ago), licensed MIT. It adds 151 tokens to every session and 1,485 once invoked, about $0.0008 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
autofix
Review and repair current local changes until they converge, or run Qwen Code Autofix issue and review workflows from GitHub Actions.
roam
Codebase comprehension via roam-code CLI. Use when exploring codebases, planning modifications, debugging failures, assessing PR risk, or checking architecture health. Triggers on: understanding project structure, pre-change safety checks, finding symbols/files, blast radius analysis, affected tests, health scoring…
vibedrift-drift-check
Use when writing or changing code in an existing repository to keep new code consistent with the repo's own conventions and avoid duplicating code that already exists. Checks a proposed function against the repo's dominant patterns and existing functions BEFORE it lands, turning drift detection into drift prevention.…
assess-findings
Triage static analysis findings, grade them against recorded decisions, and accept noise or irrelevant items.
debug
Systematic bug investigation with a five-phase framework: reproduce, narrow, diagnose, fix, verify. Supports automated regression bisect via --regression flag. Produces a structured debug report with root cause analysis, regression test, and CQ/Q self-evaluations.
mobile-flows-maestro
This skill should be used when Maestro is explicitly requested or already present and the task is to author, run, or debug iOS/Android Maestro flows; use Maestro MCP; or handle Maestro selectors, system UI, permissions, Keychain, JavaScript, waits, device state, flakiness, or CI. Evidence includes a .maestro directory…