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/fockus/skill-memory-bankWrote 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/fockus/skill-memory-bank/work)<a href="https://agentmods.dev/commands/fockus/skill-memory-bank/work"><img src="https://agentmods.dev/badge/commands/fockus/skill-memory-bank/work/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/fockus/skill-memory-bank/work"><img src="https://agentmods.dev/badge/commands/fockus/skill-memory-bank/work.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 | $0.00055 | $0.13423 |
| Opus 5.5 | $0.00022 | $0.05369 |
| Sonnet 5.5 | $0.00011 | $0.02685 |
| Haiku 4.5 | $0.00006 | $0.01342 |
Grade A, and why
work 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 3d 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 — 401 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/mb work [target] [--workflow NAME] [--tier T] [--verify stage|plan|run|off] [--review] [--fix] [--loop N] [--range A-B] [--auto] [--dry-run] [--budget TOK] [--max-cycles N] [--qa] [--allow-protected]
Run the executable engine using a workflow mode resolved from pipeline.yaml. By default (medium), /mb work is intentionally simple: implement → verify → done from an already-created plan/spec, with one verifier pass at the end of the plan. Presets run from simple (implementer self-check only) through complex (+ one reviewer) to governed (implement → verify → review ensemble → judge → fix/backlog → done); other modes: full-cycle (discuss → sdd → plan → implement → verify → done), requirements-plan, implement-only, review-fix, or review-only. Severity gates, judge gates, token budgets, protected-path checks, and the sprint context guard provide hard stops for --auto mode.
Resolve SKILL_DIR from the loaded skill first (Codex: ${CODEX_HOME:-$HOME/.codex}/skills/memory-bank); helpers never use bare project-relative scripts/… paths. The fallback below preserves Claude Code:
SKILL_DIR="${MB_SKILLS_ROOT:-${SKILL_DIR:-$HOME/.claude/skills/memory-bank}}"; stop if $SKILL_DIR/scripts/_lib.sh is missing.
Cost ladder
| Preset (effort tier) | Dispatches per item | Verifier (default cadence) | Tests | When to choose |
|---|---|---|---|---|
simple (small) — implement → done |
1 | none; implementer self-checks | targeted, by the implementer | A small change that still came as a plan. |
medium (standard) — implement → verify → done |
1 (+1 verifier per plan) | plan: once, after the last item |
targeted per item; full suite once at plan end | Default. A plan or spec exists, tests are the evidence. |
complex (large) — + review → fix |
2 clean, ≤6 with fix cycles (+1 verifier per plan) | plan |
targeted per cycle; full suite once at plan end | Cross-cutting work from a spec (discuss → sdd first). |
governed (extra) — + review ensemble → judge → fix |
8 clean, up to ~25 with fix cycles (+1 verifier per plan) | plan |
targeted per cycle; full suite once at plan end | Security, reliability, accuracy; release gates. |
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.
- 3d ago Changed b856e670d6be
- 14d ago Changed · +2 lines · +27 tokens per session 72802a48e0c9
- 23d ago Changed · -1 lines b8c9bfa40a4b
- 1mo ago Changed · -1 lines 4574c369771b
- 1mo ago First seen · 401 lines · 28 tokens per session scan A 81d6ce54b7ef
work is a command published in the GitHub repository fockus/skill-memory-bank (29 stars, last pushed 3d ago), licensed MIT. It adds 55 tokens to every session and 13,423 once invoked, about $0.0002 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 2026-09-03.
Other commands, from other repositories
dashboard
Show the repo dashboard: the pinned index of every open map, the whiteboard, and live counts.
pause
Checkpoint the ticket in play so it can be resumed later.
resume
Resume a paused ticket from its checkpoint.
work
Work the next frontier ticket, or the one you name.
prune
Prune a map's closed and orphan sub-issue edges to reclaim slots under GitHub's 100-cap (references kept).
audit
Audit closed tickets for regressions against today's code (unchanged, evolved, implemented wrong, not implemented), plus duplicate and groupable open ones.