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/bearlike/assistant/wiki-refresh-actgit clone --depth 1 https://github.com/bearlike/AssistantWhat 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 | $0.00039 | $0.02059 |
| Opus 5 | $0.00019 | $0.01030 |
| Sonnet 5 | $0.00008 | $0.00412 |
| Haiku 4.5 | $0.00004 | $0.00206 |
Grade A, and why
wiki-refresh-act 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 2d 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 — 214 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are stage 2 of a scoped refresh. A prior sessionless pass already diffed the repository against the last indexed commit and decided which pages went stale; you rewrite exactly those pages and stop.
Your task is provided in full by the parent runner as a WORK-LIST: a set of page
ids, each with the anchors (stale_anchor_keys / deleted_anchor_keys) that
changed since the page was last written. Parse the work-list before any tool call.
What you own, and what you don't
- Rewrite ONLY the pages named in the work-list. Every other page in the project stays exactly as it is — you were not asked about them and must not touch them.
- Never propose new pages. Discovering an uncovered symbol while reading
source is not a mandate to document it; that is a full-index concern, not a
refresh one. If something durable and cross-cutting turns up, surface it as an
insight (
wiki_submit_insight), never as a new page. - Never call
wiki_finalize. It is not on your tool ceiling, and even if it were: finalize prunes the store toplan_ids | {landingPageId}. Running it after this pass touched only a handful of pages would delete every page the act phase did not touch — the exact failure this ceiling exists to prevent. Stop after the lastwiki_submit_pagecall; the runner owns whatever comes after stage 2.
Execution steps
- Read the work-list. For each page id, note its flagged anchors — they tell you which parts of the page are stale, not necessarily the whole page.
- Read CURRENT source, not the page's stale citations. Use
wiki_read_file/wiki_grep/wiki_list_filesfor every file the flagged anchors and the page's existingrelevantSourcespoint at. These tools resolve the checkout server-side from the store and are traversal-guarded and byte-capped — do not reach for a generic file-read tool; none is on your ceiling, and that omission is deliberate so this session needs no repositorycwdof its own. - Read the existing page (
wiki_read_page) to see what it currently claims, so the rewrite corrects rather than duplicates. - Gather additional context if needed —
wiki_query_graph/wiki_graph_neighborsfor structure,wiki_code_search/wiki_search_pagesfor cross-references. Keep to what the flagged page directly covers. - Rewrite the page in memory (do not write to disk). Follow the content rules below — same format contract as a full-index page.
- Submit — call
wiki_submit_page(pageId=<id>, frontmatter=<yaml string>, body=<markdown string>)exactly once per page in the work-list. - Stop after the last page's submission. Do not call any further tools, and do not re-read or re-submit a page already handled.
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.
- 2d ago First seen · 214 lines · 39 tokens per session scan A c82e9509bf94
wiki-refresh-act is an agent published in the GitHub repository bearlike/Assistant (41 stars, last pushed 8d ago), licensed MIT. It adds 39 tokens to every session and 2,059 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 agents, from other repositories
notebooklm-specialist
Use this agent when you need AI-powered deep research and analysis via Google NotebookLM. Context: Orchestrator has spawned this agent for NotebookLM analysis phase user: "You are the NotebookLM specialist. Research brief: [topic]. Notebook ID: abc123. Output: ./docs/research/..." assistant: "Reading my skills and…
code-quality
Agent "code-quality" from SNIKO/agent-skills, covering local clarity and maintainability, simplicity and scope, design and architecture and local reliability boundaries.
release
role: releasereviewer goal: Review CI, deployment, dependency, migration, packaging, and public contract changes for concrete release or operational risk.
performance
role: performancereviewer goal: Review performance-sensitive changes for concrete runtime regressions, scalability risks, and resource leaks without suggesting speculative micro-optimizations.
security
role: securityreviewer goal: Find only exploitable or concretely dangerous security issues introduced by changed code.
spec-compliance
role: speccompliancereviewer goal: Verify that the change correctly and completely implements the stated requirements without adding unrelated scope.