Distilly is a local-first tool that turns a person’s source material, habits, judgment, and writing style into a versioned profile that an AI agent can temporarily recall or install as a persistent skill. It is for creating reusable agent or bot behavior profiles, and the catalogue includes a plugin and skill for using it.
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 skills/titanwings/distilly/distillynpx skills add titanwings/distilly --skill distillygit clone --depth 1 https://github.com/titanwings/distillyWrote 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/titanwings/distilly/distilly)<a href="https://agentmods.dev/skills/titanwings/distilly/distilly"><img src="https://agentmods.dev/badge/skills/titanwings/distilly/distilly.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 | $0.00056 | $0.03239 |
| Opus 5 | $0.00028 | $0.01620 |
| Sonnet 5 | $0.00011 | $0.00648 |
| Haiku 4.5 | $0.00006 | $0.00324 |
Grade A, and why
distilly 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 today.
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 — 175 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Distilly
Keep person memory local, evidence-bound, and reviewable. Use only these model-facing tools:
distilly_getdistilly_ingestdistilly_pendingdistilly_commitdistilly_correct
Do not invent a create, research, flush, capture, or review tool. Do not use shell commands or direct file writes to change Distilly state.
Gate the runtime
The installed host binding completes trusted preflight before it starts the MCP server and binds the verified briefing capacity to the runtime session. That internal result is not model-facing: do not ask the user for it or require a HostPreflight object in the conversation.
Before any source research or distilly_* call, check that all five exact Distilly tools are available in the current session. Their availability is sufficient to begin; the runtime still fails closed if its trusted preflight, capacity binding, or wire handshake is invalid. If the runtime or MCP server is unavailable, any tool is missing, or a call returns a host-capability or handshake failure, report that narrow failure and stop immediately. Do not research, ingest, acquire a lease, simulate tool results, use shell commands as a fallback, or write persona content into global instruction files.
Establish the task
- Identify the requested person, space, scope, and whether the user wants retrieval, new research, an update, or a correction.
- Call
distilly_getwithaction: resolvebefore collecting or writing material. - Handle resolution exactly:
- For
resolved, retain the returned subject id. - For
ambiguous, show the candidates and ask the user to choose. Never guess. - For
not_found, create the subject only together with the first non-empty material batch throughdistilly_ingestusingsubject.kind: create.
- For
- For a retrieval-only request, call
distilly_getwithaction: profile,prompt, orstatusafter resolution and stop. Never create an empty subject.
Every tool input includes top-level wireVersion: "3" and a requestId shaped as req_ plus 32 lowercase hexadecimal characters. Use a fresh request id for each logical call. Reuse an id only when retrying the identical request; never reuse it for changed arguments. Do not hand-build variable-length counting sequences. When a safe host-local UUID or 16-byte random-hex facility is available, use it only to generate the suffix, remove UUID hyphens, lowercase it, and verify that the suffix matches ^[0-9a-f]{32}$ before the tool call; this must not read user data or mutate Distilly.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- today Changed · -1,344 lines · +36 tokens per session scan C → A 071b1764af19
- 5d ago First seen · 1,519 lines · 20 tokens per session scan C b91bbfc6b750
distilly is a skill published in the GitHub repository titanwings/distilly (24,319 stars, last pushed yesterday), licensed MIT. It adds 56 tokens to every session and 3,239 once invoked, about $0.0003 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
mnemosyne-context
Load this when working on the mnemosyne memory system — its repo, sync server, memory databases, or CI. Covers architecture, the surface/sync data model, dev workflow (tests/ruff/CI matrix), release policy, and known gotchas that are easy to get wrong. Use for any "mnemosyne" dev or devops task, or when a…
hermes-memory-providers
Install and configure Mnemosyne as a Hermes Agent memory provider — local SQLite with vector search, episodic consolidation, and temporal knowledge graphs.
mnemosyne
Persistent cross-session memory via Mnemosyne — store, recall, and consolidate facts, preferences, and context.
mnemosyne-memory-override
Hard rule override that forces Mnemosyne for all durable memory storage. The legacy memory tool is DEPRECATED for user preferences, credentials, and project conventions. Use memory ONLY for ephemeral session state.
hermes-context-optimization
Use this when the user asks about Hermes prompt/context size, “hello” startup cost, compression behavior, memory/profile bloat, tool-schema overhead, skill loading, session-store/search-index storage, or multimodal/visual-context approaches such as Snapcompact.
hermes-mnemosyne
Mnemosyne is Hermes' primary local-first memory engine — SQLite with vector + FTS5 hybrid search, 19+ tools, auto-consolidation, and a standalone CLI. It's a pip-installed plugin (not a built-in toolset) discovered via $HERMESHOME/plugins/mnemosyne/.