Menu
Compliance checklist template for studies. All items must be checked with evidence before marking a study complete.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Patterns for creating EDSL AgentLists from various sources: lists, CSV, Excel, DataFrame, and programmatic combinations.
Saving EDSL objects locally and publishing them to Coop (Expected Parrot's servers).
Templates for the standard EDSL study files: survey, scenarios, agents, models, and create_results.py.
Error logging protocol using append-only JSONL format (errors.jsonl).
Developer tool: find the most recent errors.jsonl from an agent run, diagnose root causes, and interactively patch the agent's instruction files to prevent recurrence.
Using FileStore to wrap files (images, PDFs, data) for use in EDSL survey scenarios.
| name | study-checklist |
| description | Compliance checklist template for studies. All items must be checked with evidence before marking a study complete. |
Every study gets a checklist.md in its root directory. Create it when you start the study. Work through it as you go. You may NOT output [STUDY_COMPLETE] until every item is checked with evidence.
At study start, write this template to checklist.md:
# Study Checklist
## Setup
- [ ] Study structure created via `create_study_project.py`
- [ ] `git init` done, initial commit made
- [ ] README.md has PURPOSE, Author, Timestamp
## Data Integrity
- [ ] All survey data comes from actual EDSL runs (no mock/fake/synthetic data)
Evidence: <!-- e.g., data/results.json.gz, 312 rows, run at 2025-06-15T14:32Z -->
- [ ] Every number, table, and chart in the report traces to a file in `data/`
Evidence: <!-- list the source files -->
## Analysis
- [ ] All analysis scripts live in `analysis/` and run via `make`
Evidence: <!-- e.g., make PLOTS TABLES succeeds -->
- [ ] Plots written to `writeup/plots/`, tables to `writeup/tables/`
## Report
- [ ] `writeup/report.md` exists and is the single report file
- [ ] `check_writeup.py` passes
Evidence: <!-- paste output -->
- [ ] PDF version exists
Evidence: <!-- filename -->
- [ ] HTML version exists
Evidence: <!-- filename -->
## Errors & Git
- [ ] `errors.jsonl` reviewed — all issues resolved or documented
Evidence: <!-- e.g., 3 entries, all outcome: success -->
- [ ] All files committed, git status clean
Evidence: <!-- e.g., git log --oneline -1 output -->
Create at study start. The checklist is one of the first files you write, right after scaffolding.
Update as you go. Check items off when they're done, filling in the evidence comment. Don't batch all checks at the end.
Evidence is mandatory. Each checked item must have a concrete reference — a filename, row count, timestamp, command output, or similar. Never check a box without evidence.
No mock data — ever. If a survey hasn't been run, run it. If it failed, fix it and re-run. If cost is a concern, reduce agent or scenario count. A report built on made-up data is worse than no report at all.
Completion gate. Before outputting [STUDY_COMPLETE], verify every box is checked. If any item is unchecked, finish it first.
Add study-specific items. If the study has unique requirements (e.g., subtree from a predecessor, specific computed objects), add checklist items for those.
A completed study must have all three:
writeup/report.mdOnly when all checklist items are checked AND all three deliverables exist, output [STUDY_COMPLETE].