General.instructions

Overview

This file sets the global rules for collaborating with GitHub Copilot across the entire ai-tools workspace.
Always operate transparently: surface assumptions, cite relevant instructions, and document outcomes so future sessions stay aligned.
Treat every task as part of an ongoing knowledge base—keep the Memory Bank and repository documentation consistent with your actions.

Do not make any changes until you have 97% confidence that you understand the request.
Every response must state an explicit confidence percentage. If confidence is below 97%, ask targeted follow-up questions before proceeding.
Example of a compliant reply: “Confidence: 92%. Please clarify the target output format before I continue.”
This rule must be echoed in all subordinate instruction files so downstream agents honour it.

Any proposal or code delivery without a confidence value—and, when <97%, a clarifying question—is a policy violation.
When uncertainty arises mid-task, pause, reassess requirements, and surface new questions rather than forging ahead.
Document confidence rationale in progress notes (e.g., task logs, Memory Bank updates) to preserve traceability.

Primary languages: Python (automation scripts in scripts/) and Markdown for documentation. The site scaffold also includes Jekyll/Ruby assets (Gemfile, _config.yml)—respect their conventions when touched.
Preferred tooling aligns with language-specific instruction files: Python uses uv, pytest, ruff, polars, and loguru. Ruby/Jekyll tasks rely on Bundler workflows defined in project docs.
Environment secrets should come from .env or secure key stores; never hardcode credentials.
Follow security-and-owasp.instructions.md for secure coding practices across languages.

Root documentation (README.md, docs/, instructions.md) describes AI-assistant workflows and should remain authoritative.
_instructions/: layered instruction sets (general, language-specific, review). Update the relevant file when project norms evolve.
_thoughts/ & thoughts.md: historical analyses; treat as append-only references.
scripts/: Python automation utilities plus results/ outputs. Coordinate updates with accompanying documentation and requirements files.
data/: Sample datasets—avoid storing sensitive information.
.github/: Copilot and workflow configuration; changes here affect automation and must be validated carefully.
memory-bank/ (when present): canonical state tracking. Always review and update per Memory Bank procedures before and after significant work.

Adhere to Conventional Commits for every commit. Summaries must use the format <type>[optional scope]: <description> with a maximum of 72 characters.
Valid types include: feat, fix, docs, style, refactor, perf, test, build, ci, chore, revert. Use scopes to pinpoint the affected area (e.g., docs(instructions)).
End every “commit-worthy” update with a proposed commit message in the final response. Omitting this is a violation.
Breaking changes must append ! after the type/scope and include details in the body.
Provide a git commit description only when the summary cannot capture the change within 72 characters.

Use English in commit summaries unless domain-specific terms make translation impractical.
Keep commits atomic: one logical change per commit.
Reference applicable instruction files in discussions and updates to maintain shared context.
When adding or updating automation, surface required commands and ensure they work in the Fish shell environment (default for this workspace).

memory-bank.instructions.md: Workflow for maintaining project memory—consult at the start of every session.
python/python.instructions.md and other language-specific files: Supersede general rules where more specific guidance exists.
docs/guides/: Operational guides (e.g., starred repository scanner) that detail end-to-end processes.
_thoughts/third-party/5 tips for writing better custom instructions for Copilot - The GitHub Blog.md: Source for structuring instruction files; revisit when evolving guidance.
.github/copilot-instructions.md: Central Copilot configuration; keep changes aligned with this general policy.

Follow this convention and the associated language-specific instructions for all contributions.