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.
git clone --depth 1 https://github.com/DyanGalih/spec-kit-memory-hubWrote 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/commands/dyangalih/spec-kit-memory-hub/speckit.memory-md.capture)<a href="https://agentmods.dev/commands/dyangalih/spec-kit-memory-hub/speckit.memory-md.capture"><img src="https://agentmods.dev/badge/commands/dyangalih/spec-kit-memory-hub/speckit.memory-md.capture/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/commands/dyangalih/spec-kit-memory-hub/speckit.memory-md.capture"><img src="https://agentmods.dev/badge/commands/dyangalih/spec-kit-memory-hub/speckit.memory-md.capture.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00014 | $0.01223 |
| Opus 5 | $0.00007 | $0.00611 |
| Sonnet 5 | $0.00003 | $0.00245 |
| Haiku 4.5 | $0.00001 | $0.00122 |
Grade A, and why
speckit.memory-md.capture 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 9d 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 — 103 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Capture
Reflect on completed work and update durable memory only if needed.
Resolve configuration first. Use .specify/extensions/memory-md/config.yml when present; otherwise default to memory_root: docs/memory and specs_root: specs.
Capture is automatic based on your confidence score. Evaluate the proposed durable memory and determine your confidence (0-100%) that it is correct, durable, and non-duplicate. If your confidence is > 50%, automatically approve the capture and register it. If your confidence is <= 50%, ignore it and do not capture. However, you must always allow the user to trigger this manually and bypass the confidence check if they explicitly request a capture.
Inputs to review:
- active spec / plan / tasks
- final implementation diff or summary
- tests or validation results
- review findings (Architecture Guard, Security Review, etc.), if any
- incident or bug-fix context, if any
Durable Memory Context (Duplicate Prevention)
Before proposing new entries, check the existing memory to avoid duplicates.
- Refresh Cache: If memory search is empty or you need the latest state, call
speckit_memory_refresh_cache(scope="memory")ornpx speckit-memory refresh-memoryto restore the SQLite cache from the backup.mdfiles. - Targeted Search: Call
speckit_memory_search(query="architecture constraints boundaries decisions <topic>")for the candidate lesson topics. - Read Results: Review the search results to ensure the candidate lesson is not already captured.
- Do NOT read
.mdmemory files directly. MCP search results via SQLite are the single source of truth. The.mdfiles are only backups managed by the extension.
For each candidate lesson, require all of these:
- reusable
- non-obvious
- likely to prevent future mistakes
- evidenced by the diff, tests, review feedback, or incident analysis
- correctly scoped to durable memory rather than feature-local notes
Every new entry must answer:
- why this is durable
- what future mistake it prevents
- what evidence supports it
- where maintainers should look next
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.
- 9d ago First seen · 103 lines · 14 tokens per session scan A 4ea80eecc045
speckit.memory-md.capture is a command published in the GitHub repository DyanGalih/spec-kit-memory-hub (15 stars, last pushed 3mo ago), licensed MIT. It adds 14 tokens to every session and 1,223 once invoked, about $0.0001 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 commands, from other repositories
wiki-query
Ask questions against the wiki. Synthesizes answers from wiki pages with cross-reference citations.
wiki-req
Capture and decompose a concept into atomic, traceable wiki requirements. Clarifies ambiguous requirements, splits them into atomic pieces, and persists them as wiki/requirements/ pages with status tracking.
wiki-retro
Save an atomic insight from the current task into the wiki. Creates a single markdown file that layered recall surfaces in future sessions.
wiki-discover
Auto-discover new sources from the web. Searches based on config topics and known knowledge gaps.
wiki-init
Initialize a new LLM Wiki in the current directory. Creates the full directory structure, config, and template files.
wiki-run
Run the full wiki cycle: discover → ingest → lint. Optionally schedule for auto-updates.