using-compounded

Instructions for using compounded, a system that stores shared user preferences and manages the trust level of reusable skills. Its trust levels range from proposed and untested to verified, trusted, and autonomous.

In plain words
What is it for?
Loading general user preferences, checking or advancing skill trust levels, and exporting preferences and verified skills for use on another machine.
Why use it?
It explains which remembered preferences are loaded and when skills are safe to use automatically.

Skill for Claude CodeCodex

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/ankitkr3/compounded/using-compounded
Any agent
npx skills add ankitkr3/compounded --skill using-compounded
Clone the repo
git clone --depth 1 https://github.com/ankitkr3/compounded

Made for: Claude Code, Codex.

Per session 73 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,235 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 $0.00073 $0.01235
Opus 5 $0.00036 $0.00617
Sonnet 5 $0.00015 $0.00247
Haiku 4.5 $0.00007 $0.00123

Measured yesterday against content hash 7cd071c5b595, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

using-compounded 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 yesterday.

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/using-compounded/SKILL.md · 84 lines

How it starts

The opening of the file, as written. The whole thing — 84 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Using compounded

compounded is installed. This means three things are true:

  1. A user-global memory file lives at ~/.claude/compounded/USER.md. It is loaded into your context at session start. It is bounded — usually under 1500 characters. Use it for preferences and facts that apply across all the user's projects, not for project-specific notes (those belong in CLAUDE.md or AutoMemory's MEMORY.md).

  2. Skills you create earn trust over time. Skills do not start trusted. They start as proposals, get verified by replay, and graduate through a four-state trust ladder.

  3. There is a portable archive. The user can run /compounded:export to bundle their USER.md and verified skills into a single archive that moves between machines.

The Trust Ladder

When you (the agent) discover a useful procedure during a session — something non-trivial, repeatable, that took you more than a couple of tool calls to figure out — you may propose it as a skill. Skills then graduate through four states:

State What it means When it loads
.proposed Just authored. Untested. Never auto-loaded. Listed by /compounded:status.
.verified Replayed successfully on at least one matching task. Loads only when explicitly invoked by name.
.trusted Used 3+ times without correction. Auto-loads when relevant, asks for confirmation before applying.
.autonomous Used 10+ times, zero corrections in last 5. Auto-loads and applies without asking.

Promotions happen automatically based on usage. Demotions also happen automatically:

  • A user correction during a skill's use → demote one step
  • 3 corrections in 30 days → demote to .proposed
  • A failed verification on a .proposed skill → move to .rejected/ (recoverable, not deleted)

When to propose a skill

Use the compounded-author skill (which has the full procedure) when all of these are true:

  • The task involved more than 3 tool calls
  • The procedure is repeatable on a similar future task
  • You actually solved it (not just gave up halfway)
  • It is not redundant with an existing skill
  • It does not duplicate something native (Claude Code's bundled skills, AutoMemory, AutoDream)

Read the full file on GitHub · 84 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. yesterday First seen · 84 lines · 73 tokens per session scan A 7cd071c5b595

Subscribe to this mod's changes

using-compounded is a skill published in the GitHub repository ankitkr3/compounded (5 stars, last pushed 2mo ago), licensed MIT. It adds 73 tokens to every session and 1,235 once invoked, about $0.0004 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