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 aigorahub/elves --skill elves-grok-botgit clone --depth 1 https://github.com/aigorahub/elvesWrote 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/aigorahub/elves/elves-grok-bot)<a href="https://agentmods.dev/skills/aigorahub/elves/elves-grok-bot"><img src="https://agentmods.dev/badge/skills/aigorahub/elves/elves-grok-bot/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/aigorahub/elves/elves-grok-bot"><img src="https://agentmods.dev/badge/skills/aigorahub/elves/elves-grok-bot.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 | $0.00041 | $0.00909 |
| Opus 5 | $0.00020 | $0.00454 |
| Sonnet 5 | $0.00008 | $0.00182 |
| Haiku 4.5 | $0.00004 | $0.00091 |
Grade A, and why
elves-grok-bot 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 11d 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 — 60 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Elves (Grok Bot)
Files are memory. Chat is not. The user owns done. You own the middle.
Do not use /goal — that string is reserved for a possible future feature.
This is not Grok Build and not a coding host. No git, PR, or merge.
Stage
Create one run. Give it a unique run id. Write four files under one folder named for that id. Put the folder path, run id, and reminder id in SURVIVAL and session.json. Every wake uses that exact path. Do not invent a second folder.
- PLAN.md — mission, scope, batches, M-A# acceptance, non-negotiables
- SURVIVAL.md — start with "READ THIS FIRST." Stop gate, next action, file paths
- session.json —
status(running|stopped|blocked|complete),stop_allowed: falseuntil every M-A# has proof,run_id,reminder_id,active_fire - LEDGER.md — last item, last outcome, last record id, next action
Proof is a count, file, or record id. "The important ones are done" is not proof.
Stop gate
Stay closed until acceptance is proven, unless status is already terminal.
Not a stop: a recap, the first slice, silence, "this is a lot," a natural pause, a quiet inbox.
Hard stop only: user said stop, auth is dead, or the next write would break a rule. On a hard stop, set status to stopped or blocked before you return, then delete that exact reminder. Every later fire that sees a terminal status exits without work.
Mid-run ping: answer in a few sentences and continue.
Watchdog
One reminder only, every minute, around the clock (cron: * * * * *). Overnight unattended work is the point — the user may go to bed after kickoff. Persist the reminder id. Each fire: confirm run id and reminder id, then read survival → session → plan → ledger. If status is terminal, delete this reminder if it still exists and exit. If the gate is closed, do the next batch. Do not recap instead of working. When all M-A# are proven, set status: complete, recap once, and delete that exact reminder.
An interrupted fire (status error "Interrupted before it finished") is NOT a stop. Next fire continues from SURVIVAL/LEDGER after reconciling any in-flight item.
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.
- 11d ago First seen · 60 lines · 41 tokens per session scan A e8ee23894ce4
elves-grok-bot is a skill published in the GitHub repository aigorahub/elves (219 stars, last pushed 4d ago), licensed MIT. It adds 41 tokens to every session and 909 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 2026-08-30.
Other skills, from other repositories
opencode-session-db
Read OpenCode sessions, messages, and tool outputs directly from the SQLite database at /.local/share/opencode/. Use when asked to "read opencode sessions", "query opencode db", "find old sessions", "search session history", "read message history", "export session", "inspect opencode data", "look up past…
agent-evaluation
Evaluate GenAI agent task execution using LLM-as-judge. Produces structured scores (0-5), feedback, and improvement recommendations.
feature-workflow
Standard workflow for developing features. Follow this process for all non-trivial changes - from planning through PR merge. Ensures proper testing, review, and CI verification.
readiness-check
Verify all OpenCode plugin services are healthy and ready. Use when diagnosing plugin issues, after deployment, or when services like Whisper, TTS, Supabase, or Telegram aren't working.
authoring-a-skill
Write a new reusable skill (SKILL.md) so korgex gets better at recurring work.
browser-automation
Drive a real browser verifiably — navigate, snapshot, act BY INDEX, extract; every step provable.