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
{{ message }}
Repository navigation
Standardize repository instructions for coding agents #7
Audit the root and nested agent-instruction files in data, engineer, backtest, specs, live, diagnostic, and models. Replace the historical directory-map sidecar design with a maintained cross-agent standard based on current Codex, Claude Code, GitHub Copilot, and AGENTS.md discovery behavior.
Confirmed starting state
No managed release repository contains agents.txt; the existing convention is AGENTS.md.
The seven repositories range from no tracked instructions to 24 AGENTS.md files.
Data and Engineer have ignored local-only CLAUDE.md files; Diagnostic and Models have none on main.
Many nested files are API and file inventories with stale line counts rather than local working constraints.
Engineer exposes the sidecars through the installed package and a public get_agent_docs() API.
Acceptance criteria
Publish an ecosystem standard defining filename, root and nested scope, size limits, required content, exclusions, Claude interoperability, distribution policy, and maintenance triggers.
Record the dated audit evidence and authoritative research sources.
Define automated conformance checks that validate discovery files without testing configuration tautologies.
Create one implementation issue in each library only after the standard is reviewed.
Do not merge library instruction changes until the ecosystem standard is accepted.
The accepted standard and executable audit are implemented by milestone issue #31 in commit 1ce1547 and draft PR #41. The pull request includes a closing reference for this issue. Each stable library now has an owning semantic review issue; Backtest's machine-detected nested-guide deviations are recorded in ml4t/backtest#100.
Activity
stefan-jansen commentedon Sep 21, 2026
The accepted standard and executable audit are implemented by milestone issue #31 in commit 1ce1547 and draft PR #41. The pull request includes a closing reference for this issue. Each stable library now has an owning semantic review issue; Backtest's machine-detected nested-guide deviations are recorded in ml4t/backtest#100.
Standardize stable libraries and record closure evidence