chore(coach): trim public baseline to schema mechanism, no methodology#88
Merged
Merged
Conversation
Brian flagged (2026-07-08) that the public baseline had drifted well past "competent example" into real coaching IP: full periodization phase sequencing (base -> max strength -> conversion -> taper), a season-phase block-sequencing algorithm, event/cross-training load-accounting heuristics (freshest-legs-before, adjacent-day lightening), and ailment severity handling. All genuinely useful methodology sitting in a public AGPL repo. The trimmed baseline keeps only what a self-hosted deploy actually needs to function: tool-use mechanics, schema field usage (weekday/curves/phases), and the ability to READ every CONTEXT field (objectives, constraints, events, ailments) so the app doesn't silently ignore data the UI lets a self-hoster set. It drops the METHODOLOGY of how to reason about those fields — that now lives exclusively in the private tuned prompt (vires-ops/prompts/coach_system.txt, reconciled to the full pre-trim content in vires-ops#53, deployed via SSM as this PR's companion). Safety language (never treat/rehab a diagnosed condition, defer to a professional) stays in BOTH — never traded for IP protection. New test_baseline_prompt_stays_generic_no_proprietary_methodology guards this split going forward so the drift doesn't silently recur; the other baseline-prompt tests were narrowed to mechanism-only assertions. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Brian flagged that the public baseline prompt had drifted well past "competent example" into real coaching IP: full periodization phase sequencing, a season-phase block-sequencing algorithm, event/cross-training load-accounting heuristics, and ailment severity handling — genuinely useful methodology sitting in a public AGPL repo, with no private version actually deployed to differentiate it (see vires-ops#53).
This PR trims the committed baseline to schema mechanism only — tool-use mechanics, how to read every CONTEXT field (so a self-hosted deploy doesn't silently ignore objectives/constraints/events/ailments the UI lets a user set) — and drops the methodology of how to reason about those fields. That now lives exclusively in the private tuned prompt (
vires-ops/prompts/coach_system.txt, already reconciled to the full pre-trim content in vires-ops#53 and being deployed to SSM as this PR's companion).Safety language stays in both, unconditionally — never treat/rehab a diagnosed condition, defer to a professional. That's never traded for IP protection.
A new guard test (
test_baseline_prompt_stays_generic_no_proprietary_methodology) asserts specific methodology phrases never leak back into the public file, so this doesn't silently re-drift the way the original baseline did.Test plan
uv run pytest— full suite green (309 tests; narrowed 4 prompt-content tests to mechanism-only assertions, added the new anti-leak guard)uv run ruff check .— clean