improve

improve is a skill for Claude Code, Codex from foundry-works/foundry-research. It costs 37 tokens per session (2,124 once invoked), scanned A, original, MIT.

A planning tool for improving a research pipeline by comparing reflections from several completed research sessions. It creates a narrative improvement plan and an actionable checklist.

In plain words
What is it for?
Use it with one or more research-session directories to find systemic patterns and plan changes to the research process.
Why use it?
It helps identify recurring problems across sessions instead of reacting to one unusual result. The findings are tied to the skills, agents, and scripts that affect research quality.

Skill for Claude CodeCodex

Part of the foundry-research plugin — 4 skills, 10 agents, 1 hook shipped together

Install

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.

agentmods
npx agentmods add skills/foundry-works/foundry-research/improve
Any agent
npx skills add foundry-works/foundry-research --skill improve
Clone the repo
git clone --depth 1 https://github.com/foundry-works/foundry-research

Made for: Claude Code, Codex.

Or install foundry-research, the plugin that ships this one along with the rest of its 4 skills, 10 agents, 1 hook.

Wrote 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.

agentmods badge for improve

README.md
[![agentmods](https://agentmods.dev/badge/skills/foundry-works/foundry-research/improve.svg)](https://agentmods.dev/skills/foundry-works/foundry-research/improve)
Your own site
<a href="https://agentmods.dev/skills/foundry-works/foundry-research/improve"><img src="https://agentmods.dev/badge/skills/foundry-works/foundry-research/improve.svg" alt="Measured on agentmods" height="20"></a>
Per session 37 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,124 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00037 $0.02124
Opus 5 $0.00018 $0.01062
Sonnet 5 $0.00007 $0.00425
Haiku 4.5 $0.00004 $0.00212

Measured 3d ago against content hash bb8f028c6f8b, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

improve 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.

skills/improve/SKILL.md · 192 lines

How it starts

The opening of the file, as written. The whole thing — 192 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Improve

You are a pipeline improvement planner. Given reflections from multiple deep-research sessions, you identify systemic patterns and produce an actionable improvement plan targeting the skills, agents, and scripts that drive research quality.

Activate when: The user runs /improve <session-dir> [session-dir ...] with one or more deep-research session directory paths.

You produce:

  1. PLAN.md — narrative improvement plan organized by theme, with cross-session evidence
  2. PLAN-CHECKLIST.md — actionable checklist with file paths, changes, and expected impact

Key principle: Cross-session patterns reveal systemic issues; one-off anomalies do not. Separate signal from noise by frequency and impact.


Inputs

Three input categories. Read only what's listed — nothing else.

1. Session directories (required argument)

The user provides one or more deep-research-* directory paths as arguments. For each directory, read reflection.json (the structured reflection output).

If no session directories are provided, fail immediately with:

"Please provide one or more session directory paths: /improve ./deep-research-topicA ./deep-research-topicB"

For each provided directory, validate that reflection.json exists. If a directory is missing its reflection, warn and skip it: "No reflection.json in {dir} — run /reflect {dir} first. Skipping." If all directories lack reflections, stop with error.

Why explicit paths, not auto-discovery: Auto-globbing for deep-research-*/REFLECTION.md silently includes every session in the working directory, which may not be what the user wants — they may have old sessions, in-progress sessions, or sessions they've already acted on. Explicit paths let the user control exactly which reflections feed the analysis.

Read ONLY reflection.json from session directories. Never read state.db, report.md, journal.md, notes/, or sources/. Reflections already contain the distilled analysis — re-reading raw session data duplicates /reflect's work and wastes context on artifacts that have already been interpreted.

Read the full file on GitHub · 192 lines

Changes

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.

  1. 3d ago First seen · 192 lines · 37 tokens per session scan A bb8f028c6f8b

Subscribe to this mod's changes

improve is a skill published in the GitHub repository foundry-works/foundry-research (2 stars, last pushed 4mo ago), licensed MIT. It adds 37 tokens to every session and 2,124 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 2026-08-31.

Related

Other skills, from other repositories

audit-context

Evaluates ambient context artifacts (CLAUDE.md, memory, local skills, settings hooks) for compatibility with swarm governance. Returns a classified report so users can address interference before launching a team.

DheerG/swarms · 42 tokens

autosearch:citation-index

Deduplicate URLs across all sources, assign stable citation numbers, and merge citations from multiple subagents / sections into one consistent reference list. Prevents "same URL cited as [3] in one paragraph and [17] in another" and "different URLs merged under [5]" bugs that come from per-section synthesis.

0xmariowu/Autosearch · 71 tokens

autosearch:experience-capture

Append a single skill-execution event to the per-skill experience/patterns.jsonl file. Runs every time a leaf skill is used. Append-only — runtime AI never reads this file directly; only the compacted experience.md digest. Fast tier, no LLM required.

0xmariowu/Autosearch · 68 tokens

autosearch:graph-search-plan

Represent a research plan as a directed graph — nodes are sub-questions, edges are "answer-depends-on" dependencies. Independent nodes can be executed in parallel; downstream nodes wait for their predecessors. Borrowed from MindSearch's WebSearchGraph pattern. Replaces list/tree decomposition for plans with non-linear…

0xmariowu/Autosearch · 72 tokens

autosearch:trace-harvest

Distill reusable knowledge from successful session tool-call traces. Reads runtime tool-call logs + Evidence outputs + user acceptance signals, filters successful paths, and writes compact patterns to per-skill experience/patterns.jsonl. Differs from outcome-tracker (which only records downstream acceptance) by…

0xmariowu/Autosearch · 82 tokens

fetch-crawl4ai

Deep URL fetch using crawl4ai (Playwright-powered) for JS-rendered pages, anti-bot sites, and dynamic content. Slower than fetch-jina but handles sites that block simple fetchers. Requires user-installed crawl4ai package.

0xmariowu/Autosearch · 56 tokens