Skip to content

feat: values-based autonomy framework and channel-agnostic triage - #2

Closed
cchen362 wants to merge 2 commits into
ben4mn:mainfrom
cchen362:feat/autonomy-framework
Closed

feat: values-based autonomy framework and channel-agnostic triage#2
cchen362 wants to merge 2 commits into
ben4mn:mainfrom
cchen362:feat/autonomy-framework

Conversation

@cchen362

@cchen362 cchen362 commented Mar 3, 2026

Copy link
Copy Markdown
Contributor

Summary

  • Add AUTONOMY_FRAMEWORK identity/values/systems prompt injected into all LLM calls
  • Add platform-aware context builder for macOS/Windows/Linux
  • Make heartbeat triage triggers channel-agnostic with _build_channel_guidance()
  • Update default system prompt to values-aligned version

Context

Implementation Plan 002. Builds on PR #1 (cross-platform). Replaces hardcoded iMessage references in triage with dynamic channel guidance. Merge PR #1 first.

Test plan

  • Verify AUTONOMY_FRAMEWORK appears in system prompt
  • Verify platform context shows correct OS
  • Verify triage triggers use channel_guidance placeholder

cchen362 added 2 commits March 4, 2026 04:45
Replace os.uname() calls with sys.platform checks for Windows compat.
Add platform-aware shell execution (cmd.exe on Windows, bash on Unix).
Add PowerShell equivalents of setup/start/restart scripts.
Add run.py for Windows-safe uvicorn startup (SelectorEventLoop).
Unpin strict dependency versions for Python 3.13 compatibility.
Add AUTONOMY_FRAMEWORK identity/values/systems prompt injected into all
LLM calls. Add platform-aware context builder for macOS/Windows/Linux.
Make heartbeat triage triggers channel-agnostic with _build_channel_guidance()
replacing hardcoded iMessage references. Update default system prompt to
values-aligned version.
@ben4mn

ben4mn commented Mar 4, 2026

Copy link
Copy Markdown
Owner

Thanks for this — interesting idea, but closing for now. A few concerns:

  1. Personality changes without migration — this changes Edward's core system prompt and identity for existing users with no opt-in or migration path
  2. Permanent token cost — ~380 extra tokens injected into every single LLM call adds up significantly
  3. Self-evolution encouragement — "You can modify your own code" is a significant autonomy escalation that needs careful consideration
  4. Forward references — references NotebookLM capabilities before PR feat: integrate Google NotebookLM as a skill with 12 tools #3 adds them, which would confuse the LLM
  5. Channel-agnostic triage refactor is a good change but should be submitted separately

If you'd like to revisit the triage refactor as a standalone PR, that part would be welcome!

@ben4mn ben4mn closed this Mar 4, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants