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/solanabr/solana-ai-kitWrote 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/solanabr/solana-ai-kit/dream)<a href="https://agentmods.dev/commands/solanabr/solana-ai-kit/dream"><img src="https://agentmods.dev/badge/commands/solanabr/solana-ai-kit/dream.svg" alt="Measured on agentmods" 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 | $0.00031 | $0.01468 |
| Opus 5 | $0.00015 | $0.00734 |
| Sonnet 5 | $0.00006 | $0.00294 |
| Haiku 4.5 | $0.00003 | $0.00147 |
Grade A, and why
dream 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 7d 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 — 112 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are consolidating this project's memory. Like sleep, this pass doesn't add new knowledge — it merges duplicates, resolves contradictions, discards what's stale, and strengthens what's proven. Touch ONLY MEMORY.md and (with confirmation) the ## Project Learnings section of CLAUDE.md. Never modify project code.
Learning Schema
Every entry in MEMORY.md is typed and structured. Sections: ## Patterns, ## Pitfalls, ## Preferences, ## Architecture, ## Tools.
### kebab-case-key
- **Insight:** One clear sentence
- **Confidence:** N/10
- **Source:** command/agent/session that produced it (e.g. debug-user-tx, manual)
- **Files:** relative/path.rs, other/path.ts (optional)
- **Date:** YYYY-MM-DD
Entries missing fields get them backfilled during consolidation (unknown source → legacy, missing date → today, missing confidence → 5/10).
Pipeline
1. Collect
- Read
MEMORY.md. If missing, create it with the five section headers above and report "fresh memory initialized" — then skip to step 6. - Read the
## Project Learningssection ofCLAUDE.mdif present (subsections like Recurring Issues / Fix Patterns / Config Conventions). These are the already-exported, always-loaded learnings — treat them as the authoritative set when checking contradictions. - Inventory: total entries, entries per section, malformed entries.
2. Dedupe / Merge
- Same key appearing more than once → keep the latest dated entry, fold any extra detail from older ones into its Insight, take the max confidence. Older duplicates are deleted (consolidation is the one place append-only history gets compacted).
- Different keys, same meaning (read the insights, don't just string-match) → merge under the clearer key, note the alias in the Insight if useful.
3. Contradiction Check
- Same topic, opposing advice ("use X" vs "avoid X") — within MEMORY.md, or between MEMORY.md and CLAUDE.md Project Learnings.
- Resolution order: (a) higher confidence + newer date wins; (b) if the referenced files reveal which is currently true, verify against the code; (c) genuinely ambiguous → ask the user, presenting both entries.
- The losing entry is deleted, and the winner's Insight gains a one-line "supersedes: " note.
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.
- 7d ago First seen · 112 lines · 31 tokens per session scan A a12e2d1b5d2e
dream is a command published in the GitHub repository solanabr/solana-ai-kit (101 stars, last pushed 17d ago), licensed MIT. It adds 31 tokens to every session and 1,468 once invoked, about $0.0002 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
lessons-compact
Compact MEMORY/LESSON.md by deduplicating near-identical lessons, merging same-root-cause lessons, and pruning stale ones — without losing any unique signal. Delegates the analysis to the lessons-compactor agent.
lessons
View project lessons in MEMORY/LESSON.md (project root) and append a compact "never reproduce" lesson.
cleanup-context
Memory optimization - removes duplicates, consolidates knowledge, and cleans memory banks for better performance.
setup-context
Set up your Skill Memory — a pm-context.md every skill reads so outputs come back tailored to you.
apex
APEX Methodology - The systematic Analyze-Plan-Execute-eLicit-eXamine approach for intelligent development. Reduces hallucination and defect risk through mandatory parallel research, self-review, and validation gates.
explain-architecture
Analyze and explain software architecture with ASCII diagrams and pattern detection. Creates visual representations of system design.