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 commands/fockus/skill-memory-bank/adrgit clone --depth 1 https://github.com/fockus/skill-memory-bankWhat 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 | $0.00010 | $0.00815 |
| Opus 5 | $0.00005 | $0.00407 |
| Sonnet 5 | $0.00002 | $0.00163 |
| Haiku 4.5 | $0.00001 | $0.00081 |
Grade A, and why
adr 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 2d 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 — 78 lines — stays where its author put it; the contents beside it link to each section on GitHub.
ADR: $ARGUMENTS
0. Validate arguments
If $ARGUMENTS is empty, stop and ask the user for the decision title. Do not proceed with an empty title.
1. When to record an ADR (3-gate)
Offer an ADR only when all three are true (from grill-with-docs):
- Hard to reverse — the cost of changing your mind later is meaningful.
- Surprising without context — a future reader will look at the code and wonder "why on earth did they do it this way?"
- The result of a real trade-off — there were genuine alternatives and you picked one for specific reasons.
If any gate fails, skip the ADR: an easy-to-reverse decision will just be reversed; an unsurprising one needs no explanation; a no-alternative one records nothing beyond "we did the obvious thing." When the user explicitly asks for an ADR, honour it — but note which gate is weak.
2. Context
- Read
./.memory-bank/backlog.md— existing ADRs live in the## Architectural decisions (ADR)section. - If no
## Architectural decisionssection exists, create it at the bottom ofbacklog.mdunder a---divider. - Study the relevant part of the codebase so the decision has real grounding.
- Make sure this decision (or a close variant) has not already been recorded or rejected — search with
grep -i "<keyword>" .memory-bank/backlog.md.
3. Determine the next ADR number
grep -oE 'ADR-[0-9]+' .memory-bank/backlog.md | sort -V | tail -1
# Take the numeric part, add 1. Zero-pad to 3 digits (ADR-001, ADR-002, ...).
# If no existing ADR → start from ADR-001.
Numbering is monotonic — never reuse an ID, even if an ADR was later rejected or replaced.
4. Draft the decision
Show the user a draft before writing. Required parts:
- Context — what problem are we solving, what constraints exist, why this decision is needed now
- Decision — what exactly we decided to do
- Alternatives — which options were considered and why each was rejected
- Consequences — what changes because of this decision, what trade-offs, what becomes easier / harder
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.
- 2d ago First seen · 78 lines · 10 tokens per session scan A c202d9e21a2f
adr is a command published in the GitHub repository fockus/skill-memory-bank (25 stars, last pushed 1mo ago), licensed MIT. It adds 10 tokens to every session and 815 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
story-implement
Implement a story through the Implement Phase. Stories must be fully designed with chunks before implementation.
planning
Feature roadmap and planning. Manage initiatives, concepts, open questions, and story creation from planning.
story-new
Create a new story. It lands in the backlog until assigned to a cycle.
backlog
The project layer: milestones, epics, and user stories with refusing controllers — spec-seeded, sprint-ready.
cycle-complete
Complete a cycle. Triggers reflection if pending learnings, then archives.
cycle-design
Design a cycle — create new cycles with planned stories, detail existing planning cycles, or quick-sketch a roadmap. Detects planning docs in .craft/planning/ and sources the cycle from them when relevant.