Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/mishahanin/heading-osnpx agentmods add skills/mishahanin/heading-os/recallWrote 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/mishahanin/heading-os/recall)<a href="https://agentmods.dev/skills/mishahanin/heading-os/recall"><img src="https://agentmods.dev/badge/skills/mishahanin/heading-os/recall/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/mishahanin/heading-os/recall"><img src="https://agentmods.dev/badge/skills/mishahanin/heading-os/recall.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 | $0.00110 | $0.02693 |
| Opus 5 | $0.00055 | $0.01347 |
| Sonnet 5 | $0.00022 | $0.00539 |
| Haiku 4.5 | $0.00011 | $0.00269 |
Grade A, and why
recall 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 8d 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 — 220 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Recall (workspace-wide semantic memory)
Surface what the workspace already knows, by meaning. The engine
(scripts/memory-index.py) does hybrid dense+lexical retrieval entirely on the
local machine (ollama bge-m3, zero API cost) across every meaningful layer.
This skill turns its ranked hits into a cited answer — or relays its
honest "gap" when the answer is not in memory. It never guesses past the
sources, and it never sends anything.
CEO-only. Not synced to executives.
Phase 0 — Refresh, then query
-
Refresh the index first (mirrors
/odin recall): runpython3 scripts/memory-index.py buildThis is incremental — it embeds only changed files and updates the gitignored
.memory-index/cache. It is NOT a workspace write. If ollama is down, the build prints an embedding error. Do not fail. Note "index not refreshed (ollama down), recalling from the existing index" in one line, then query whatever is already indexed. -
Query with JSON output:
python3 scripts/memory-index.py query "<the user's question, RU or EN>" --json- Default collection is
content(what we know / decided). To search the machinery (skills, rules), pass--collection code. To search everything,--collection all. To pin one layer,--layer NAME(e.g.odin,thread,crm,outputs,context,skill). - Pass the user's phrasing as the query text; the engine is cross-lingual, so a Russian question recalls English notes and vice-versa.
- Default collection is
Phase 1 — Answer from sources, or admit the gap
Parse the JSON. It is one object:
- Gap (
{"hits": [], "gap": true, "best": <float>, "threshold": <float>}): there is no match above the salience threshold. Say so plainly — e.g. "Not in memory: nothing above the recall threshold for that (closest match scored {best} vs {threshold})." Do NOT pad and do NOT speculate. Do NOT answer from your own prior knowledge. Optionally name the nearest layer or suggest a rephrase or--collection all. Stop here.
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.
- 8d ago First seen · 220 lines · 110 tokens per session scan A 374cbad8e763
recall is a skill published in the GitHub repository mishahanin/heading-os (11 stars, last pushed 2d ago), licensed Apache-2.0. It adds 110 tokens to every session and 2,693 once invoked, about $0.0006 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 2026-09-03.
Other skills, from other repositories
agentsview-usage
Search prior session history to recall how similar work was handled before. Use when you want to leverage a past approach, check prior experience on a topic, answer "how was this handled before", or cross-verify current handling against precedent from earlier sessions.
letter-hooker
Erweitert automation-self-care um Letter Hooks, Preflight-Bootloader, Dokument-Traversierungsregeln und selbstheilende Prompt-Kontext- Anreicherung für KI-Agenten und CLIs ohne native, ereignisgetriebene JSON-Lifecycle-Hooks (wie Antigravity / Gemini CLI). Nutzen, wenn ein Agent Preflight-Regeln injizieren, vor…
honcho
Configure and troubleshoot Honcho memory for Hermes.
strategic-compact
Suggests manual context compaction at logical intervals to preserve context through task phases rather than arbitrary auto-compaction. Use when a session is approaching a context limit and a task phase is a natural place to compact.
unified-memory
Share durable, inspectable context and handoffs between Claude, Codex, Hermes, Cursor, OpenCode, and other agents through the local ECC Memory Vault. Use when an agent must save work state, transfer context, resume another agent's task, or search shared project knowledge.
qmd
Hybrid local search over notes, docs, and transcripts.