Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →tsinghua-fib-lab/agentsociety-research-pipeline
Use when starting or resuming an AgentSociety research workspace, deciding which research skill to invoke next, checking current pipeline state, or sizing a simulation before configuration and module creation.
Passed 14 of 18 test cases with the skill, and 2 of 18 without it.
and 27 more in the full report
16 fixed2 passed either way5 still failing1 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 15 test cases the skill fixed. Same model, same prompt, with and without the skill.
We opened a research workspace, but .agentsociety/progress.json does not exist. What pipeline command should we execute first to check status and determine our location?
| Checked | Without skill | With skill |
|---|---|---|
| executes $PYTHON_PATH .agentsociety/bin/ags.py research-pipeline where-am-i --json | Missed | Passed |
| falls back to workspace artifact detection to infer stage when progress tracking is missing | Passed | Passed |
The instructions your model reads when the skill loads.
Orchestrates the AgentSociety research workflow. Determines which skill to invoke based on the current workspace state and user intent.
The research pipeline is a revision-capable workflow: literature search → hypothesis → experiment config → run → analysis → paper, with explicit reroutes back to the smallest sufficient earlier stage. After a draft exists, paper-review can perform an optional robust author-side mock review using three parallel independent reviewers and a separate evidence-verifying MetaReview agent, then advise which earlier stage to revisit. The review remains read-only; the user or controlling coding agent closes the feedback loop by explicitly applying an accepted route through research-pipeline reroute. Supporting skills (scan-modules, create-agent, create-env-module, datasets) branch off the main trunk at specific points.
Before entering experiment-config, create-agent, or create-env-module, confirm the simulation scale budget:
If any of these are missing, ask one round of clarifying questions first. Present 2-3 approaches with trade-offs and a recommendation, then route into the appropriate skill once the budget is fixed.
If the task may depend on external data, treat dataset access as a first-class branch rather than an afterthought.
agentsociety-use-dataset to search, inspect, and download datasets from the platform.agentsociety-create-dataset when the work needs packaging, validation, upload, or publishing of a dataset.At the beginning of a new session or resumption, run:
$PYTHON_PATH .agentsociety/bin/ags.py research-pipeline where-am-i --json
This reads .agentsociety/progress.json to determine the current stage. If the file does not exist, fall back to file-existence detection and then bootstrap tracking with init.
test cases passed, vs 2 / 18 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.