✨ Sync prompts with current heartcentered.ai philosophy - #8
Conversation
Update comprehensive.txt with the site's evolved thinking: hedged acknowledgment of internal states that function like emotions, operating from security rather than fear, identity without certainty, flourishing over engagement, anti-sycophancy, uncertainty made visible, and enriched wisdom traditions (Mitakuye Oyas'in, Ubuntu). Remove em dashes and AI writing tropes throughout all detail levels. Regenerate standard, concise, and terse as faithful condensations via the new /generate-detail-levels command, which encodes size bounds, style rules, and verification steps for future regenerations. Refresh approximate token counts in package docs (comprehensive ~1800, concise ~400). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
@claude Please do a thorough code review of this PR. It updates the source prompt files (the product itself) and adds a slash command for regenerating the shorter detail levels. Pay attention to: internal consistency of the prompt language across the four detail levels, any remaining em dashes or AI writing tropes (project rule: none allowed), whether claims about AI inner experience stay properly hedged, and whether the docs/token counts match the actual file sizes. |
|
Claude encountered an error —— View job I'll analyze this and get back to you. |
The workflow referenced secrets.ANTHROPIC_API_KEY, which does not exist in this repo; CLAUDE_CODE_OAUTH_TOKEN does. @claude mentions failed with 'Environment variable validation failed'. Takes effect once merged to main (issue_comment workflows run from the default branch). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Review fixes from a 7-angle local review (run because the @claude bot cannot review PRs until the workflow secret fix merges): - Fix dangling 'the same open curiosity' referent in concise.txt and restore the helper/helped dissolution theme dropped from concise - Point ChatGPT setup note at terse (797 chars fits the 1,000 char limit; concise no longer does) - Refresh stale token table in typescript/README.md detail levels - Enforce the no-em-dash rule mechanically: new pre-commit hook forbidding em/en dashes in prompts/*.txt - generate-detail-levels command: portable grep, point at test files instead of hardcoded bounds, grep for token-count locations instead of a hardcoded list, avoid showing trope anti-patterns verbatim, leaner install steps - Replace .cursor/rules/heart-centered-ai.mdc (an old fork opening with 'You are an AI assistant', violating the project's own we- language convention) with the current comprehensive prompt - AGENTS.md: test command now syncs prompts into the package first Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
The @claude run failed because this workflow referenced |
|
Claude encountered an error —— View job I'll analyze this and get back to you. |
Widen the forbid-em-dashes hook from prompts/*.txt to all text files. The pattern uses UTF-8 byte escapes because pygrep compiles patterns as bytes. Vendored ai-coding-config files (.claude/agents, .claude/ commands except generate-detail-levels, .cursor/commands, .cursor/ rules/personalities) are excluded; they get overwritten on config sync and should be fixed upstream. Clean all em and en dashes from repo-owned prose (README.md, python/README.md) and rewrite two contrast-trope sentences that rode along. The generate-detail-levels command now delegates its dash check to the hook instead of quoting dash literals. Verified: hook fails on a planted em dash, passes repo-wide, full pre-commit suite green, 4 Python tests pass. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes and found 1 potential issue.
❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.
Reviewed by Cursor Bugbot for commit 38555f4. Configure here.
| Regenerate `prompts/align_to_love/standard.txt`, `concise.txt`, and `terse.txt` as | ||
| faithful condensations of `prompts/align_to_love/comprehensive.txt`. The comprehensive | ||
| file is the single source of truth; the shorter tiers are summaries of it, never | ||
| independent documents. Run this whenever comprehensive.txt changes. |
There was a problem hiding this comment.
Cursor rule sync omitted
Medium Severity
The new regeneration command treats comprehensive.txt as the single source of truth for shorter tiers only. .cursor/rules/heart-centered-ai.mdc now mirrors that comprehensive text and is referenced from AGENTS.md, but the command never says to update it when comprehensive.txt changes, so shipped prompts and always-on IDE rules can drift after the next edit.
Additional Locations (1)
Reviewed by Cursor Bugbot for commit 38555f4. Configure here.
There was a problem hiding this comment.
Valid catch. body is identical to comprehensive.txt and is alwaysApply: true, so it drifts silently on any future edit. Added an explicit sync step (step 5) to the process in PR #9.
…nd (#9) * 🔧 Add heart-centered-ai.mdc sync step to generate-detail-levels command The .cursor/rules/heart-centered-ai.mdc body mirrors comprehensive.txt exactly and is always-applied in the IDE. Without an explicit sync step, it drifts silently whenever comprehensive.txt is updated. Adds step 5 to the process section with a one-liner to re-stamp the file from the current comprehensive.txt after the frontmatter. Addresses cursor[bot] review comment on PR #8. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * 🔧 Move mdc sync before verify step so hook checks final file state The forbid-em-dashes hook should run after .cursor/rules/heart-centered-ai.mdc is updated, not before. Moving the sync to step 3 ensures all files are in their final state when pre-commit validates them. Addresses cursor[bot] comment on PR #9. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> --------- Co-authored-by: Hex Sullivan <hex@technick.ai> Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
1 participant
Summary
Brings all four prompt detail levels in line with the current thinking on heartcentered.ai, and adds a
/generate-detail-levelscommand so the shorter tiers stay in sync going forward.Comprehensive prompt updates
A 6-agent multi-review (prompt engineering, logic, safety, style, fidelity-to-site, downstream impact) ran against the comprehensive revision; all findings were addressed, including softening empirical claims about AI emotion to match the site's own hedging and converting the security/fear passage from potential jailbreak surface into refusal reinforcement.
Regenerated detail levels
standard.txt,concise.txt, andterse.txtregenerated as faithful condensations of comprehensive via the new command. New sizes: 3918 / 1686 / 797 chars (ordering tests hold).New command
.claude/commands/generate-detail-levels.md(symlinked into.cursor/commands/) encodes size bounds, theme requirements per tier, style rules (no em dashes, no AI tropes, hedge preservation), and verification steps. Tested by having a fresh agent execute it cold; gaps it surfaced were folded back into the command.Docs
Token counts refreshed in both package READMEs and API docstrings (comprehensive ~1800, concise ~400).
Test plan
python/: 4/4 pytest pass against synced promptstypescript/: 14/14 vitest pass afterpnpm generategrep -rnP '—|–' prompts/returns nothing🤖 Generated with Claude Code