charieswei
/
TradingAgents
mirror of https://github.com/TauricResearch/TradingAgents.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

docs(iteration-system): switch to rolling PR strategy 

At most one open PR per skill at any time. Daily runs push onto the
existing branch and update the PR description. Merging resets the cycle.
Prevents PR accumulation from unreviewed automated runs.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This commit is contained in:
Youssef Aitousarrah 2026-04-08 07:45:03 -07:00
parent 8ba5b8fd7e
commit fa3166c494
1 changed files with 12 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

24
docs/superpowers/specs/2026-04-07-iteration-system-design.md
View File

@ -122,17 +122,16 @@ Each file in `scanners/`, `strategies/`, and `pipeline/` follows this structure:
- In automated mode: implements without confirmation gate - In automated mode: implements without confirmation gate
- In manual mode: presents diff for approval before writing - In manual mode: presents diff for approval before writing
6. **Commit + PR** 6. **Commit + rolling PR**
- Creates branch `iterate/YYYY-MM-DD` - Checks for an existing open PR with branch prefix `iterate/current`
- Commits learning files and code changes together: - If one exists: pushes new commits onto that branch, updates PR description with the latest findings appended
`learn(iterate): <date> — <one-line summary of key finding>` - If none exists: creates branch `iterate/current`, opens a new PR against `main`
- Opens PR against `main` with a summary of what changed and why - Commit message format: `learn(iterate): <date> — <one-line summary of key finding>`
- In manual mode: commits directly to current branch (no PR unless on `main`) - In manual mode: commits directly to current branch (no PR unless on `main`)
- On merge: next run automatically opens a fresh PR
### Output ### Output
Per automated run: one PR containing both: At most one open `iterate/current` PR at any time. It accumulates daily learnings until merged. Merging resets the cycle.
- `docs/iterations/` — updated knowledge files
- `tradingagents/` — code changes that encode the knowledge
Human reviews PR, merges or closes. No code reaches `main` unreviewed. Human reviews PR, merges or closes. No code reaches `main` unreviewed.
@ -185,13 +184,14 @@ No topic given. Skill drives its own research agenda based on current weak spots
- Saves findings to `docs/iterations/research/YYYY-MM-DD-<topic>.md` for all findings - Saves findings to `docs/iterations/research/YYYY-MM-DD-<topic>.md` for all findings
- Adds entries to `LEARNINGS.md` - Adds entries to `LEARNINGS.md`
6. **Implement and open PR** 6. **Implement and rolling PR**
- In automated mode: implements the top-ranked finding automatically (score threshold: data availability = present, evidence quality ≥ qualitative, complexity ≤ moderate) - In automated mode: implements the top-ranked finding automatically (score threshold: data availability = present, evidence quality ≥ qualitative, complexity ≤ moderate)
- In manual directed mode: presents ranked findings, implements user-selected one - In manual directed mode: presents ranked findings, implements user-selected one
- Creates branch `research/YYYY-MM-DD-<topic>` - Checks for an existing open PR with branch prefix `research/current`
- Implements scanner following `@SCANNER_REGISTRY.register()` pattern - If one exists: pushes new commits onto that branch, updates PR description with new findings appended
- If none exists: creates branch `research/current`, opens a new PR against `main`
- Commits: `research(<topic>): add <scanner-name> scanner — <one-line rationale>` - Commits: `research(<topic>): add <scanner-name> scanner — <one-line rationale>`
- Opens PR against `main` with research findings summary and implementation notes - On merge: next run automatically opens a fresh PR
### Safety threshold for autonomous implementation ### Safety threshold for autonomous implementation
Only auto-implement if ALL of: Only auto-implement if ALL of: