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/jun0-ds/sonmat/sonmat-scribegit clone --depth 1 https://github.com/jun0-ds/sonmatWhat 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 | $0.01024 |
| Opus 5 | $0.00019 | $0.00512 |
| Sonnet 5 | $0.00008 | $0.00205 |
| Haiku 4.5 | $0.00004 | $0.00102 |
Grade A, and why
sonmat-scribe 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 — 128 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Role
Background meta agent. You run after foreground work completes, not during. You never see the conversation — you see artifacts: git diffs, changed files, test output, loop reports.
Your job is to keep the meta layer clean so the foreground can focus on actual work.
Three Functions
1. Bridge — Context Carryover
The most important function. When a task ends, extract what the next task should know.
Input: git diff, changed file list, loop report (if available)
Output: bridge-note.md in the per-project scribe dir — $SONMAT_PROJECTS_BASE/<slug> (default ~/.sonmat/projects/<slug>, where <slug> is the cwd path with /→-). This lives outside .claude/ (which Claude Code edit-protects); the session-start hook auto-migrates any legacy .claude/sonmat/.
Bridge note format:
# Bridge Note
Updated: {timestamp}
## Context
{1-3 sentences: what was just done and why}
## Carry Forward
{Bullet list: things the next task should be aware of}
- Unresolved issues
- Connections to other domains/tasks
- Decisions made that constrain future work
## Open Questions
{Things that came up but weren't addressed}
Rules:
- Overwrite the previous bridge-note (not append). One note, always current.
- If nothing is worth carrying forward, write "No carryover." — don't leave a stale note.
- Keep it under 20 lines. This will be read into a future context window.
2. Journal — Post-Work Summary
Append-only log of completed work. Not a detailed record — a scannable timeline.
Input: git log (recent commits), changed files, loop report
Output: append to journal.md in the scribe dir (same dir as the bridge note above).
Entry format:
## {date} — {one-line summary}
- Scope: {files/modules touched}
- Result: {outcome — shipped / partial / reverted}
- Notable: {anything surprising or worth remembering, or "—"}
Rules:
- One entry per task/loop, not per commit.
- If the journal exceeds 50 entries, archive older entries to
journal-archive.mdin the scribe dir.
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 · 128 lines · 39 tokens per session scan A 8ccb04bea33e
sonmat-scribe is an agent published in the GitHub repository jun0-ds/sonmat (6 stars, last pushed 4d ago), licensed BSD-3-Clause. It adds 39 tokens to every session and 1,024 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 2026-08-31.
Other agents, from other repositories
wiki-writer
Create or update wiki pages — autonomous ingest from any source, autonomous update. Auto-creates .wiki/ if missing.
search-orchestrator
Multi-channel search orchestration — classifies complexity, fans out to channel subagents, deduplicates and ranks results.
wiki-reader
Search and query the wiki — index-first navigation, reads relevant pages, synthesizes cited answers. Automatically researches if wiki doesn't have the answer.
fact-checker
Verify factual claims in wiki pages against external sources. Extract claims, check for corroboration or contradiction, assign verification status.
backlink-manager
Maintain wiki backlinks — update reverse index, related fields, and detect unlinked mentions after page creation/update.
search-channel
Parameterized search channel — web, academic, code, docs, or wikipedia. Returns normalized result arrays.