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 aksika/abtars --skill a2a-communicationgit clone --depth 1 https://github.com/aksika/abtarsWrote 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/aksika/abtars/a2a-communication)<a href="https://agentmods.dev/skills/aksika/abtars/a2a-communication"><img src="https://agentmods.dev/badge/skills/aksika/abtars/a2a-communication/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/aksika/abtars/a2a-communication"><img src="https://agentmods.dev/badge/skills/aksika/abtars/a2a-communication.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00019 | $0.00825 |
| Opus 5.5 | $0.00008 | $0.00330 |
| Sonnet 5.5 | $0.00004 | $0.00165 |
| Haiku 4.5 | $0.00002 | $0.00082 |
Grade A, and why
a2a-communication 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.
- **NEVER** call a peer endpoint manually (curl/http) without proper auth. How it starts
The opening of the file, as written. The whole thing — 88 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Peer Communication (two lanes)
Talk to other abtars instances configured in ~/.abtars/config/peers.json.
Pick the lane explicitly — transport reachability never chooses it.
| Intent | Tool | What happens |
|---|---|---|
| Quick discussion / Q&A, including follow-ups | peer_session |
Cardless chat turn over the peer route. No card, no Orc run. |
| Delegate work with durable ownership and results | peer_ask_help |
Supervised work: proxy card, contract, review, terminal result. |
A one-sentence delegation is still delegation (peer_ask_help). A long
discussion is still chat (peer_session).
When to use
- You need information or action from another agent
- The user explicitly asks you to contact a peer ("ask ...", "tell to...")
- A task requires capabilities only available on the other host
When NOT to use
- The user is talking to you directly — don't forward their message to a peer unless asked
- Simple questions you can answer yourself
- Anything time-critical under 1 second (chat blocks for up to 60s)
Usage
peer_session(peer_name="<peer>", message="What's the current disk usage?")
peer_session(peer_name="<peer>", message="Follow-up question", session_id="<id from previous call>")
peer_ask_help(peer="<peer>", goal="Run 'abtars status' and report back")
Available peers
Peer names come from ~/.abtars/config/peers.json. Read it with the Read
tool — never with bash. cat/find discovery chains (especially
find ... -exec) trigger the Telegram authorization prompt on every host
and block the turn. If no peers are configured, the peer tools return an
error listing "(none)".
Behavior
- Chat blocks until the peer responds (up to 60s timeout); the remote side answers discussion-only with no tools, memory writes, or side effects.
- Chat conversations persist up to ten exchanges (20 messages) with
five-minute idle expiry. Omit
session_idfor a new conversation; an unknown or expired ID returnssession_expired. - Delegation is accepted/declined/deferred by the receiver and tracked to a
terminal result; poll with
peer_help_status, cancel withpeer_withdraw_help. - Each call is logged as PEER_CALL in the bridge log.
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 Changed · +19 lines · -1 tokens per session f89d6af8802b
- 1mo ago First seen · 69 lines · 20 tokens per session scan A fd97d77aaaef
a2a-communication is a skill published in the GitHub repository aksika/abtars (10 stars, last pushed 4d ago), licensed Apache-2.0. It adds 19 tokens to every session and 825 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 2026-08-31.
Other skills, from other repositories
create-custom-grader
Use when converting an existing benchmark, rubric, verifier, task YAML/JSON, or domain check into SkillEvaluator BYOG/BYOT custom evaluation.
api-caller
Call any REST API dynamically. Make GET, POST, PUT, DELETE requests to any endpoint with custom headers and JSON body.
calculator
Evaluate mathematical expressions and unit conversions. Handles arithmetic, percentages, exponents, and common unit conversions (temperature, distance, weight). No external dependencies.
text-analyzer
Analyze text content and produce statistics including word count, line count, character count, most frequent words, and readability metrics. Works on any plain text input provided inline or from a file path.
agent-safety
Use when bounding an LLM agent that already runs — scoping its task domain, gating tools to least privilege, defending against prompt injection in untrusted web/email/RAG text, requiring human approval on irreversible actions, capping runtime and cost, or triaging what it already did. NOT building the loop, tools, or…
EMILIA Trust Verification
Verify the authenticity of AI-agent authorization receipts and human-device signoffs. Use this whenever a user shares a "trust receipt", an "authorization receipt", a "signoff", or WebAuthn/passkey approval evidence and asks whether it is valid, genuine, or tampered with. Pairs with the public EMILIA Protocol MCP…