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 instructions/ognistik/ursus/agents-mdgit clone --depth 1 https://github.com/ognistik/ursusWrote 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/instructions/ognistik/ursus/agents-md)<a href="https://agentmods.dev/instructions/ognistik/ursus/agents-md"><img src="https://agentmods.dev/badge/instructions/ognistik/ursus/agents-md.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 | $0.00732 | $0.00732 |
| Opus 5 | $0.00366 | $0.00366 |
| Sonnet 5 | $0.00146 | $0.00146 |
| Haiku 4.5 | $0.00073 | $0.00073 |
Grade A, and why
ursus AGENTS.md 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 — 49 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Read docs/MAINTAINER_NOTES.md before making substantial changes.
Project Identity
- Current MCP server name exposed to clients is
ursus. - Do not rename package targets, binary names, config paths, or MCP tool names unless the user explicitly asks for that migration.
Product Rules
- This is a Bear-specific local MCP server for macOS.
- Reads come from Bear's local SQLite database.
- Writes go through Bear's x-callback-url scheme.
- Do not write directly to Bear's SQLite database.
- Do not rebuild the old AppleScript/Shortcuts bridge into the new runtime.
- Do not turn the MCP surface into one giant
bear_actiontool.
Important User Preferences
- Batch operations are important. Prefer non-empty
operations: [...]inputs for mutation tools that support batching. - The old Alter schema and AppleScript bridge are reference material only and must not be modified.
bear_replace_contentshould compute the full note text locally, then commit through Bear's full replacement path.- That replacement flow may be used for title changes when the leading title markdown is changed.
- Discovery tools should return compact note summaries, while
bear_get_notesremains the full-note fetch. Inbox-note discovery is driven by configured inbox tags. - Mutation tools should return compact receipts, not full note bodies, unless the user explicitly asks for content.
- Bear's own trash/restore flows are intentionally excluded from the MCP surface for now. Backup restore remains available through
bear_restore_notes.
Current Technical Direction
- Config and templates currently live under
~/Library/Application Support/Ursus. - Create-note templates are currently single-file
template.md, not separate header/footer files. - The current read adapter is real and queries the installed Bear database schema.
- The current write adapter launches Bear URLs and uses best-effort DB polling for mutation receipts.
- The MCP server should run as a single instance; startup now takes a process lock to prevent stale concurrent servers.
- Runtime artifacts live under
~/Library/Application Support/Ursus. - Debug tracing is written under
~/Library/Application Support/Ursus/Logs/debug.logwith size-based retention. - Sparkle update UI must remain owned by the app executable. Long-running
ursus mcp/ursus bridge serveuse a backgroundSPUUpdaterscheduler and only hand off to foreground Sparkle UI when an update is actually available. - Background Sparkle checks must stay silent on "no update" and on updater errors. User-initiated checks may show Sparkle's normal "up to date" UI.
- Do not make ordinary one-shot CLI commands participate in scheduled Sparkle checks or advance
SULastCheckTime. - Do not add another helper app for update presentation. Preserve the current Sparkle-only foreground mode in the main app executable and the separation between bridge process ownership and update UI ownership.
- If you touch Sparkle, bridge lifecycle, launcher behavior, or multi-process app/CLI handoff, update
docs/ARCHITECTURE.mdanddocs/MAINTAINER_NOTES.mdin the same change and preserve the documented invariants.
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 · 49 lines · 732 tokens per session scan A b661a8bf554b
ursus AGENTS.md is an instructions file published in the GitHub repository ognistik/ursus (18 stars, last pushed 29d ago), licensed MIT. It adds 732 tokens to every session, about $0.0037 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 instructions, from other repositories
LibrAIum AGENTS.md
AGENTS.md instructions for nel-neru/LibrAIum, covering agents.md, language, what libraium is, commands and architecture.
mcp-obsidian-cli CLAUDE.md
Instructions for stonematt/mcp-obsidian-cli, covering claude.md, project, constraints, commands and run the server.
open-zk-kb AGENTS.md
Instructions for mrosnerr/open-zk-kb, covering open-zk-kb — agent knowledge base, commands, testing, structure and .agents/.
llm-wiki-mcp AGENTS.md
AGENTS.md instructions for jaronlu/llm-wiki-mcp, covering repository guidelines, project structure & module organization, architecture & design source, build, test, and development commands and coding style & naming conventions.
hive-memory CLAUDE.md
Instructions for moonx010/hive-memory, covering hive memory — unified memory system for ai agents, project overview, tech stack, architecture and v3 (current) — sqlite + entity model.
obsidian-mcp-server AGENTS.md
AGENTS.md instructions for nbaradar/obsidian-mcp-server, covering obsidian mcp server — agents.md, purpose, implementation plan, resources (planned) and tools (current).