You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The original criterion Tested by running an agent through testnet flow is
unverified and now tracked explicitly in #39. This issue covers the completed
skill content, installation and public hosting; moving the check does not claim
that a live agent session has passed.
Implementation merged in #50. SKILL.md now explicitly requires the user's own Privy app and local secrets, external master approvals, resumable funding stages, exchange-result reconciliation, and verified rotation/revocation. Tests validate frontmatter, published/embedded source consistency, the error catalog, and every documented command input through actual CLI schemas offline.
Keeping the final criterion open: a coding agent must read/use the installed skill in an actual dedicated testnet session. Record the agent/session ID, commit, lifecycle outcome and redacted event/audit evidence. The deterministic nightly script and mocked tests are not that agent session. Setup and boundaries are in docs/agent-workflow.md and docs/testnet.md. The GitHub testnet environment currently has no QA secrets.
Verified HTTP 200 with text/markdown and an exact byte match to packages/agent-cli/SKILL.md, including YAML frontmatter. The docs generator keeps it synchronized, llms.txt advertises the URL, and the skill guide has a working download link. Stale-content/export checks and browser smoke cover the endpoint on both local and Pages base paths. Deployment passed: https://github.com/bloxwap/agent-cli/actions/runs/38976838059 .
Public hosting is complete. The actual coding-agent testnet session acceptance remains pending the dedicated QA setup already recorded above.
Activity
joeblau commentedon Oct 16, 2026
Implementation merged in #50. SKILL.md now explicitly requires the user's own Privy app and local secrets, external master approvals, resumable funding stages, exchange-result reconciliation, and verified rotation/revocation. Tests validate frontmatter, published/embedded source consistency, the error catalog, and every documented command input through actual CLI schemas offline.
Keeping the final criterion open: a coding agent must read/use the installed skill in an actual dedicated testnet session. Record the agent/session ID, commit, lifecycle outcome and redacted event/audit evidence. The deterministic nightly script and mocked tests are not that agent session. Setup and boundaries are in docs/agent-workflow.md and docs/testnet.md. The GitHub testnet environment currently has no QA secrets.
joeblau commentedon Oct 16, 2026
The canonical raw skill is now live at https://bloxwap.github.io/agent-cli/SKILL.md via #52.
Verified HTTP 200 with text/markdown and an exact byte match to packages/agent-cli/SKILL.md, including YAML frontmatter. The docs generator keeps it synchronized, llms.txt advertises the URL, and the skill guide has a working download link. Stale-content/export checks and browser smoke cover the endpoint on both local and Pages base paths. Deployment passed: https://github.com/bloxwap/agent-cli/actions/runs/38976838059 .
Public hosting is complete. The actual coding-agent testnet session acceptance remains pending the dedicated QA setup already recorded above.