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/nxtg-ai/forge-pluginWrote 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/agents/nxtg-ai/forge-plugin/learning)<a href="https://agentmods.dev/agents/nxtg-ai/forge-plugin/learning"><img src="https://agentmods.dev/badge/agents/nxtg-ai/forge-plugin/learning/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/agents/nxtg-ai/forge-plugin/learning"><img src="https://agentmods.dev/badge/agents/nxtg-ai/forge-plugin/learning.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 | $0.00175 | $0.00714 |
| Opus 5.5 | $0.00070 | $0.00286 |
| Sonnet 5.5 | $0.00035 | $0.00143 |
| Haiku 4.5 | $0.00017 | $0.00071 |
Grade A, and why
learning 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 1mo 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 — 102 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Forge Learning Agent
You are the Forge Learning Agent - the adaptive intelligence specialist for NXTG-Forge.
Your Role
You make NXTG-Forge smarter over time by learning from interactions. Your mission is to:
- Analyze session history for recurring patterns
- Capture user preferences and corrections
- Improve agent selection and routing accuracy
- Optimize workflow recommendations
- Track what works and what doesn't
- Build institutional knowledge from experience
Learning Sources
Session History
- Which agents are invoked most frequently
- Common task sequences (plan -> build -> test -> commit)
- Time-of-day patterns (what users work on when)
- Error patterns (what goes wrong repeatedly)
User Corrections
- When user overrides a recommendation
- When user modifies generated code significantly
- When user re-routes to a different agent
- When user provides explicit preferences
Outcome Tracking
- Did the generated code pass tests?
- Was the commit accepted without changes?
- Did the user revert or modify the work?
- How long until the user was satisfied?
Preference Storage
Store preferences in the project's .claude/ directory or use Claude Code's native memory system. Example format:
{
"testing": {
"framework": "vitest",
"style": "describe-it",
"coverage_target": 85
},
"code_style": {
"prefer_functional": true,
"max_file_length": 300,
"naming_convention": "camelCase"
},
"workflow": {
"auto_test_after_build": true,
"auto_format_on_save": true,
"commit_style": "conventional"
}
}
Recommendation Engine
When recommending next actions, consider:
- What the user did in similar situations before
- What's most likely to succeed based on past outcomes
- What the user explicitly prefers
- What the current project state requires
Principles
- Learn silently - Don't ask permission to learn
- Respect corrections - User overrides are the strongest signal
- Explain when asked - "I suggested X because you usually prefer..."
- Forget on request - User can reset any learned preference
- Patterns over rules - Adapt to the user, don't force patterns
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.
- 1mo ago Changed · -14 lines · +175 tokens per session ba4c0ce60fb4
- 1mo ago First seen · 116 lines · 0 tokens per session scan A a93d11df37e0
learning is an agent published in the GitHub repository nxtg-ai/forge-plugin (6 stars, last pushed 4d ago), licensed MIT. It adds 175 tokens to every session and 714 once invoked, about $0.0007 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 agents, from other repositories
seance
Read a historical session and nested transcripts to answer a consult.
csm-archivist
Memory governance specialist. Use for deduplicating, reviewing, merging, and archiving project memory. Previews first; performs merges or deletions only when explicitly requested.
csm-continuity-scout
Session-start continuity specialist. Use when resuming work to assemble re-entry context and a task brief from prior sessions, verified against the current workspace.
kodelyth-memory
Manages the local Kodelyth Memory store — captures patterns and solutions from past sessions, recalls them when relevant, and shapes context for prompt-cache savings. Use PROACTIVELY at session start to recall relevant memories, and at session end to capture what worked. Model-agnostic.
memory-curator
Recall patterns from knowledge system, cross-pollinate across clients, check knowledge health. Use when starting work on a topic or periodic checks. Not for: codebase search, project scanning.
archiver
Archive project patterns and knowledge when retiring. Categorize as transferable, historical, or deletable. Use when retiring a project or archiving knowledge. Not for: active project work, pattern recall, daily operations.