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 skills add 23blocks-OS/ai-maestro-plugins --skill memory-searchgit clone --depth 1 https://github.com/23blocks-OS/ai-maestro-pluginsWrote 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/23blocks-os/ai-maestro-plugins/memory-search)<a href="https://agentmods.dev/skills/23blocks-os/ai-maestro-plugins/memory-search"><img src="https://agentmods.dev/badge/skills/23blocks-os/ai-maestro-plugins/memory-search/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/23blocks-os/ai-maestro-plugins/memory-search"><img src="https://agentmods.dev/badge/skills/23blocks-os/ai-maestro-plugins/memory-search.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 2 findings, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Excessive Agency · line 24 Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
- medium Data Exfiltration · line 149 Code scans file system directories looking for sensitive files. This could be reconnaissance for credential theft.Fix: Remove unnecessary filesystem scanning. If file access is needed, use explicit, scoped paths. Avoid reading ~/.ssh, ~/.aws, or credential directories.
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 | $0.00104 | $0.00693 |
| Opus 5.5 | $0.00042 | $0.00277 |
| Sonnet 5.5 | $0.00021 | $0.00139 |
| Haiku 4.5 | $0.00010 | $0.00069 |
Grade A, and why
memory-search 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 24d 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 — 43 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Memory Search
Your context window holds only this session. AI Maestro keeps your long-term memory, rebuilt nightly from your own past conversations (including ones Claude Code has already deleted). It is yours alone; other agents have their own.
- Memory cards: one-sentence statements of durable knowledge (decisions and why, facts about systems, preferences, lessons), each backed by the passages it came from. "Seen in N sessions" is weight: more sessions, more established.
- Entity graph: the things you work with (services, hosts, buckets, repos, files, people) and directed relations between them, such as
runs_on,depends_on,stores_data_in,deploys_to,breaks. A relation marked (no longer) was reported as ended in a later session.
Read what is already injected
AI Maestro adds ## Memory: blocks to your context: standing decisions and preferences at session start, and with each prompt, what you know about the entities it names. Read those before searching; search when they don't cover the question. Memory can be out of date, so confirm against the current files or system before acting on it.
Commands
memory-search.sh --about "<entity>" # relations (current first, ended marked) + cards that mention it
memory-search.sh "<query>" # memory cards first, then matching conversation passages
memory-search.sh "<query>" --mode semantic # related concepts, different wording
memory-search.sh "<query>" --mode term # exact text
memory-search.sh "<query>" --mode symbol # code identifiers (function, class, constant names)
memory-search.sh "<query>" --role user # only what the user said (or: --role assistant)
memory-search.sh "<query>" --limit 5 # default 10
Default mode is hybrid, which suits most searches.
Before deploying, moving, deleting, migrating or reconfiguring something, run --about on it: it shows what runs on it, stores data in it and depends on it, which is the part of a change that is easy to miss from the files alone. Files show the current state; memory shows why it is that way and what broke last time.
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.
- 24d ago Changed · -161 lines · -6 tokens per session 321f17606a13
- 1mo ago Changed · +44 lines · +20 tokens per session b2dab1a1cf23
- 1mo ago First seen · 160 lines · 90 tokens per session scan A 44b6638d6fe2
memory-search is a skill published in the GitHub repository 23blocks-OS/ai-maestro-plugins (11 stars, last pushed 4d ago), licensed MIT. It adds 104 tokens to every session and 693 once invoked, about $0.0004 per session on Opus 5.5. 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 skills, from other repositories
memory-search
Runs read-only SQL over the local EvoFlux session database (evoflux.db) for structured analysis that ordinary recall cannot answer, such as aggregating tool failures across sessions, filtering turns by role, agent, or time range, reconstructing a complete execution chain, or checking a remembered claim against what a…
learner
Post-debugging knowledge extraction — captures non-obvious, codebase-specific learnings that pass quality gates. Invoke after resolving tricky bugs or discovering surprising behavior.
context-management
Strategies for managing AI agent context windows including optimization, summarization, retrieval-augmented generation, progressive disclosure, and pruning. Use when the user is hitting context limits, building long-running agents, implementing RAG, optimizing token usage, or designing systems that need to manage…
knowledge-persistence
Patterns for learning from corrections, capturing instincts, extracting stable patterns, and building cross-session knowledge. Use when the user corrects you, when you discover a recurring pattern, or when insights should persist beyond the current session.
memory-patterns
Patterns for structuring MEMORY.md and topic-based memory files for Claude Code persistent memory. Use when the user asks about remembering things across sessions, wants to organize project knowledge, or needs guidance on what to save vs not save.
context-engineering
Use this skill for optimizing AI context windows. Covers context sizing, prioritization, and avoiding context overflow.