Setup James Skills

Reconcile only the current repository's root AGENTS.md so coding agents can discover:

• tech-preferences — preferred, experimental and avoided technologies;
• engineering-principles — design for local reasoning;
• verifiable-trust — independent verification and bounded AI autonomy.

This is a mutating setup skill. Explicit invocation authorises a bounded AGENTS.md change, not cross-repo changes, skill installation, or unrelated refactoring.

Reconciliation procedure

1. Determine the repository root and read AGENTS.md if it exists. Inspect existing references and documented project exceptions; preserve repository-specific instructions and unrelated content.
2. Check whether all three skills are installed/discoverable in the current agent environment. Report missing skills; do not claim installation or install silently. References may still be written to assist future setup.
3. Find the exact marker pair <!-- james-skills:start --> and <!-- james-skills:end -->. If both appear exactly once in the right order, replace only the contents of that managed block with the canonical block below if needed. If markers are incomplete, duplicated or malformed, preserve the file and report the ambiguity rather than guessing.
4. If no managed block exists, find any equivalent existing personal-guidance section. Update that section in place where this is unambiguous to avoid duplication. Otherwise append the canonical block separated by blank lines; create root AGENTS.md if absent.
5. Keep all unrelated instructions untouched; avoid whole-file reformatting. Follow repository line-ending conventions and ensure a final newline.
6. Re-read and diff the result. Confirm exactly one managed block and each of the three skill references once within the block, with the instruction to read their full contents before implementation and pass that requirement to workers. Re-running must make no changes. Report whether created/updated/already correct, and any unavailable skills.