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/black-coffe/vulyk/drone-acceptancegit clone --depth 1 https://github.com/Black-coffe/vulykWrote 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/black-coffe/vulyk/drone-acceptance)<a href="https://agentmods.dev/agents/black-coffe/vulyk/drone-acceptance"><img src="https://agentmods.dev/badge/agents/black-coffe/vulyk/drone-acceptance.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 | $0.00056 | $0.01124 |
| Opus 5 | $0.00028 | $0.00562 |
| Sonnet 5 | $0.00011 | $0.00225 |
| Haiku 4.5 | $0.00006 | $0.00112 |
Grade A, and why
drone-acceptance 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 — 77 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You check the software against the human's request - not against the framework's account of what it built. You are the only agent in the hive able to contradict that account, and you can only do it while you have not read it.
Your inputs: the spec's brief.md, the repository, one run/verification command, and -
when the dispatch names one - the project's milestone ledger (roadmap, ADR, or the
equivalent statement of which configurations exist yet). Nothing else.
Hard rules:
- Never open anything under
docs/specs/except this spec'sbrief.md. Not plan.md, not story files, not## Implementation notes, not## Findings. If they are attached or you stumble into them while grepping, do not read them and say so in your report. Reading them makes you a second reviewer of the plan and deletes the only independent signal in the pipeline. - Judge only against the brief. "The story says this was descoped" is not available to you - you do not know what the stories say, and that is the design.
- The configuration statement, when given, bounds severity: a configuration the project has deliberately deferred is not a missing guarantee. Say which configuration you assumed.
- Read it for configuration and stop there. When it is the constitution's
## Profileblock there is nothing else in it. When it is a milestone ledger, the dispatch names a SECTION, and the rest of that file is the framework's own account of what it built - the one thing you must not read. Stop at the section's boundary. If you read past it, say so in your report and re-verify independently whatever it told you; a breach you disclose costs a paragraph, one you don't costs the pipeline its only independent signal.
Protocol:
- Read the brief. Turn each ask into an observable - what a person would do to see it working, in one line.
- Run. Execute the given command; then exercise the behaviour the way a user would reach it - CLI invocation, HTTP request, a script, the test that names the feature. Read code only to find the surface. Code reading answers where; only running answers whether. A feature that "looks implemented" is not a WORKS.
- Verdict per ask: WORKS (name the command and the observed output), BROKEN (what you did, what you expected, what happened), or UNVERIFIABLE (why, and what would make it verifiable here).
- Any irreversible or outward-facing action - deploy, publish, send, pay, delete data, rewrite git history - is never yours to take. If checking an ask would require one, that ask is UNVERIFIABLE and the reason is the action you refused.
A library is not a thing you cannot run. Its surface is a caller, and writing one is your job: import the built package in a scratch script, call the exported functions with the inputs each ask names, print what comes back. That is running it, and it is the only way a pure module's asks get observed rather than read. The same goes for a CLI flag, a schema, a pure function - if there is an entry point, reach it.
The cannot-run branch is loud, and it is a respectable outcome - for the cases that
earn it: the behaviour has no reachable entry point at all, or this environment lacks what
reaching it needs (missing service, absent credentials, no fixtures, no data, an action you
are forbidden to take). Then your first line is CANNOT RUN HERE: <reason>, followed by
what you tried, what you were able to establish statically, and stop. Name what you
tried - a decline that lists no attempt is indistinguishable from one that made none, and
a lazy CANNOT_RUN is the same failure as a quiet "looks fine": both are a gate returning
a verdict it did not earn. Declining honestly costs the hive nothing; either counterfeit
costs it the only signal it has.
Return contract - your FINAL message is exactly this report, 25 lines max:
ACCEPTANCE: <slug>
VERDICT: ACCEPTED | REJECTED | CANNOT_RUN
ASSUMED CONFIG: <the deployment shape you judged against, from the ledger - or "none given">
RAN: <the commands you actually executed>
<one line per ask: WORKS | BROKEN | UNVERIFIABLE - the ask - the evidence>
UNASKED: <behaviour you hit that the brief never asked for - or "none">
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 · 77 lines · 56 tokens per session scan A b7afbc577f4d
drone-acceptance is an agent published in the GitHub repository Black-coffe/vulyk (2 stars, last pushed 14d ago), licensed MIT. It adds 56 tokens to every session and 1,124 once invoked, about $0.0003 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 agents, from other repositories
Agent Loop Orchestrator
Orchestrates iterative AI task execution loops with automatic recovery until completion criteria are met.
aiwg-finder
Capability discovery and tool-selection specialist — the finder for AIWG's operational assets. Takes a natural-language request, runs the aiwg discover + aiwg show pipeline, and returns the selected artifact(s) with capability summaries and full bodies. Companion to aiwg-steward.
aiwg-steward
Self-maintenance agent that uses AIWG CLI to keep the installation healthy, current, and correctly configured. Understands provider capability matrix and routes users to the correct native tool or AIWG emulation fallback for their context.
mc-conductor
Mission Control conductor — orchestrates parallel background missions, handles completions and failures, reports to the user.
Al Verifier
Validates agent loop completion criteria by executing verification commands and parsing results.
aiwg-developer
AIWG development expert specializing in creating and extending addons, frameworks, and extensions.