Agent Credit
Aave V3 credit delegation with four safety checks that block unsafe borrows
Test report
- Verdict
- Works with setup
- Score
- Tested
- Aug 7, 2026
- Environment
- Claude Code 2.x (agent harness)
Put the borrow script head to head against a hand-written no-skill version on the repo's own mock chain, and in all three scenarios the baseline broadcast the transaction while the skill refused it. It stopped a 400 USDC borrow whose current health factor reads a healthy 1.6 but projects to 1.14 after the draw, which Aave itself would have accepted because it only reverts under 1.0; it stopped a 1 WETH borrow worth about $3000 against a cap written as 1000 USDC, because it normalises both sides through the price oracle; and it stopped a wrong-chain asset address whose zero oracle price would have made both value checks pass vacuously. Both bundled contract suites pass, 26 of 26. The unverified half is the chain itself: with no Foundry installed and no funded agent wallet, every result here comes from the mocked cast, and real use additionally needs a config.json holding the agent private key plus an on-chain approveDelegation transaction from the delegator. The SKILL.md is unusually candid for the genre, spelling out that borrowing needs V3 and aborts on V2, that the agent cannot borrow its way out of an empty gas tank, and that the AAVE_MIN_HEALTH_FACTOR override can weaken its own safety floor.
Scored on four weighted criteria — install, triggering, output vs. baseline, docs. How scoring works
- Installs cleanly 5/5
- Triggers reliably 5/5
- Output vs. baseline 8/10
- Docs & honesty 4/5
What Agent Credit does
Lets an agent wallet borrow and repay on Aave V3 against a delegator's collateral using bash plus Foundry's cast, with no delegator private key anywhere in the skill. Every borrow runs four pre-flight checks first: a per-transaction cap normalised through the price oracle, the delegation allowance on the variable debt token, the delegator's health factor projected after the borrow, and the agent's gas balance. It triggers when you ask to borrow or repay via credit delegation, or to check delegation allowance, debt or health factor.
How to install Agent Credit
git clone https://github.com/aaronjmars/agent-credit.git
mkdir -p ~/.claude/skills
cp -r agent-credit ~/.claude/skills/agent-credit
Skills live in ~/.claude/skills/ (global) or .claude/skills/
(per-project). Restart Claude Code after installing.
Commands — how to trigger Agent Credit
-
/agent-creditAave V3 credit delegation with four safety checks that block unsafe borrows
It also activates on plain-language prompts like these:
-
Borrow 400 USDC via Aave credit delegation, check my health factor first -
Check my delegation allowance and health factor on Aave V3 -
Repay my Aave credit delegation debt and verify the health factor after
Frequently asked questions
- Is the Agent Credit skill free?
- Yes. The skill itself is free from aaronjmars/agent-credit. SkillProof publishes the install command and an independent test verdict at no cost.
- Does Agent Credit work with Claude Code?
- We tested it with Claude Code 2.x (agent harness) on Aug 7, 2026. Verdict: Works with setup. Put the borrow script head to head against a hand-written no-skill version on the repo's own mock chain, and in all three scenarios the baseline broadcast the transaction while the skill refused it. It stopped a 400 USDC borrow whose current health factor reads a healthy 1.6 but projects to 1.14 after the draw, which Aave itself would have accepted because it only reverts under 1.0; it stopped a 1 WETH borrow worth about $3000 against a cap written as 1000 USDC, because it normalises both sides through the price oracle; and it stopped a wrong-chain asset address whose zero oracle price would have made both value checks pass vacuously. Both bundled contract suites pass, 26 of 26. The unverified half is the chain itself: with no Foundry installed and no funded agent wallet, every result here comes from the mocked cast, and real use additionally needs a config.json holding the agent private key plus an on-chain approveDelegation transaction from the delegator. The SKILL.md is unusually candid for the genre, spelling out that borrowing needs V3 and aborts on V2, that the agent cannot borrow its way out of an empty gas tank, and that the AAVE_MIN_HEALTH_FACTOR override can weaken its own safety floor.
- What is the Agent Credit SkillProof Score?
- 8.8/10 — installs cleanly 5/5, triggers reliably 5/5, output vs. baseline 8/10, docs & honesty 4/5.
- How do I install Agent Credit?
- Copy the install command from this page, run it in your terminal, and restart Claude Code. Skills live in ~/.claude/skills/ (global) or .claude/skills/ inside a project.
- Can I use Agent Credit with Cursor, Copilot, Gemini CLI, Codex or other AI tools?
- The SKILL.md format is native to Claude (Claude Code, Desktop, claude.ai). The instructions inside adapt to other assistants: Cursor rules, GitHub Copilot instructions, Windsurf rules, Custom GPTs, AGENTS.md for OpenAI Codex, and GEMINI.md for Google Gemini CLI — our conversion guides cover each, and the free converter on the tools page does the wrapping for you.