How do I install the Cacheforge skill?

Run npx skillmds@latest add modbender/cacheforge in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Cacheforge safe to use?

This skill has not completed SkillMD's automated safety review yet. Independent scanners report: SkillSpector: PASS, Skill Scanner: PASS. Capability flags: executes scripts. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Cacheforge?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Cacheforge free to use?

Yes. Installing skills from SkillMD is free. This skill is licensed under MIT.

Who published Cacheforge?

modbender (@modbender) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Cacheforge by modbender — Agent Skill

Name: Cacheforge
Rating: 4.666666666666667 (3 reviews)
Author: modbender

File contents

Purpose

cacheforge is the primary entrypoint skill. Install this one skill first.

On first use, it bootstraps companion skills if missing:

cacheforge-setup
cacheforge-ops
cacheforge-stats

Then it routes the user request:

setup/onboarding -> cacheforge-setup
billing/upstream/keys -> cacheforge-ops
usage/savings dashboard -> cacheforge-stats

CacheForge Positioning

CacheForge is an OpenAI-compatible gateway for agent workflows. It can reduce wasted spend and improve repeat-turn performance (results vary by provider/workload).

Vault Mode (Pro) is for tool-heavy agents. Verify results in the CacheForge dashboard.

Bootstrap Workflow (Required)

Before routing to companion skills, run:

bash "{baseDir}/scripts/bootstrap-companions.sh"

If bootstrap fails because clawhub is missing, tell the user to install companions manually:

for s in cacheforge-setup cacheforge-ops cacheforge-stats; do clawhub install "$s"; done

Routing Rules

Use these routes after bootstrap:

Setup / first-time onboarding:
- python3 "{baseDir}/../cacheforge-setup/setup.py" provision ...
- python3 "{baseDir}/../cacheforge-setup/setup.py" openclaw-apply --set-default
- python3 "{baseDir}/../cacheforge-setup/setup.py" validate
Account ops:
- python3 "{baseDir}/../cacheforge-ops/ops.py" balance
- python3 "{baseDir}/../cacheforge-ops/ops.py" topup --amount 10 --method stripe
- python3 "{baseDir}/../cacheforge-ops/ops.py" topup --amount 10 --method crypto
- python3 "{baseDir}/../cacheforge-ops/ops.py" upstream ...
- python3 "{baseDir}/../cacheforge-ops/ops.py" keys ...
Stats / savings:
- python3 "{baseDir}/../cacheforge-stats/dashboard.py" dashboard
- python3 "{baseDir}/../cacheforge-stats/dashboard.py" usage --window 7d
- python3 "{baseDir}/../cacheforge-stats/dashboard.py" breakdown --by model

Onboarding Guardrails

Always enforce this order for new users:

Register + verify email (if required by deployment).
Create tenant API key (cf_...).
Configure upstream provider (openrouter, anthropic, or custom).
Apply OpenClaw provider config (with backup).
Top up credits (minimum top-up is typically $10).
Run a validation request and check dashboard telemetry.

Public Copy Rules

Do not use hard numeric savings claims in public-facing output unless linked to a reproducible benchmark.
Do not reveal Vault internals.
Keep copy outcome-focused and include "results vary by provider/workload" when relevant.

---name: cacheforgedescription: CacheForge primary skill — bootstrap onboarding + ops + stats for the OpenAI-compatible token optimization gateway. Cut your LLM bill by up to 30% (results vary by provider/workload).license: MIT--- ## Purpose `cacheforge` is the primary entrypoint skill. Install this one skill first. On first use, it bootstraps companion skills if missing: - `cacheforge-setup` - `cacheforge-ops` - `cacheforge-stats` Then it routes the user request: - setup/onboarding -> `cacheforge-setup` - billing/upstream/keys -> `cacheforge-ops` - usage/savings dashboard -> `cacheforge-stats` ## CacheForge Positioning CacheForge is an OpenAI-compatible gateway for agent workflows. It can reduce wasted spend and improve repeat-turn performance (results vary by provider/workload). Vault Mode (Pro) is for tool-heavy agents. Verify results in the CacheForge dashboard. ## Bootstrap Workflow (Required) Before routing to companion skills, run: ```bash bash "{baseDir}/scripts/bootstrap-companions.sh" ``` If bootstrap fails because `clawhub` is missing, tell the user to install companions manually: ```bash for s in cacheforge-setup cacheforge-ops cacheforge-stats; do clawhub install "$s"; done ``` ## Routing Rules Use these routes after bootstrap: - Setup / first-time onboarding: - `python3 "{baseDir}/../cacheforge-setup/setup.py" provision ...` - `python3 "{baseDir}/../cacheforge-setup/setup.py" openclaw-apply --set-default` - `python3 "{baseDir}/../cacheforge-setup/setup.py" validate` - Account ops: - `python3 "{baseDir}/../cacheforge-ops/ops.py" balance` - `python3 "{baseDir}/../cacheforge-ops/ops.py" topup --amount 10 --method stripe` - `python3 "{baseDir}/../cacheforge-ops/ops.py" topup --amount 10 --method crypto` - `python3 "{baseDir}/../cacheforge-ops/ops.py" upstream ...` - `python3 "{baseDir}/../cacheforge-ops/ops.py" keys ...` - Stats / savings: - `python3 "{baseDir}/../cacheforge-stats/dashboard.py" dashboard` - `python3 "{baseDir}/../cacheforge-stats/dashboard.py" usage --window 7d` - `python3 "{baseDir}/../cacheforge-stats/dashboard.py" breakdown --by model` ## Onboarding Guardrails Always enforce this order for new users: 1. Register + verify email (if required by deployment). 2. Create tenant API key (`cf_...`). 3. Configure upstream provider (`openrouter`, `anthropic`, or `custom`). 4. Apply OpenClaw provider config (with backup). 5. Top up credits (minimum top-up is typically `$10`). 6. Run a validation request and check dashboard telemetry. ## Public Copy Rules - Do not use hard numeric savings claims in public-facing output unless linked to a reproducible benchmark. - Do not reveal Vault internals. - Keep copy outcome-focused and include "results vary by provider/workload" when relevant.

Frequently asked questions

Cacheforge