[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7DKSW7MJQ8A88TJC1H293PB":3,"skill-same:01M7DKSW7MJQ8A88TJC1H293PB":310},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":16,"sourceHtmlUrl":17,"categories":18,"keywords":19,"publishedAt":20,"linkOnly":21,"markdownContent":22,"markdownBody":23,"frontMatter":24,"refName":25,"commitSha":26,"skillDirectory":27,"forksCount":14,"repositoryHtmlUrl":28,"repositoryHomepageUrl":6,"ownerHtmlUrl":29,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":30,"install":31,"resources":44},"01M7DKSW7MJQ8A88TJC1H293PB","aer-statspai",null,"Use when aer-identification has fixed the design, after methodology choice and before aer-robustness or aer-tables-figures, to run an AER-track analysis with StatsPAI — the agent-native Python engine and MCP server for causal inference, robustness, sensitivity, and publication-ready table export.","vc999999999\u002Fskillflux-catalog","vc999999999","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F104196942?v=4","01M7DKFQK2JRAY2C9EMD7FHBWG","jiangbing","01M7DKFQKQA5E0BWZ1WQY4G9WA",0,"MIT","https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: aer-statspai\ndescription: Use when aer-identification has fixed the design, after methodology choice and before aer-robustness or aer-tables-figures, to run an AER-track analysis with StatsPAI — the agent-native Python engine and MCP server for causal inference, robustness, sensitivity, and publication-ready table export.\n---\n\n# AER StatsPAI\n\n## Overview\n\n`aer-statspai` is the **implementation engine** option for this stack. Where\n`aer-identification` and `aer-robustness` decide *which* estimator and *which*\ndiagnostics a referee will demand, this skill is about *running* them — through\n[StatsPAI](https:\u002F\u002Fgithub.com\u002Fbrycewang-stanford\u002FStatsPAI), an open-source,\nagent-native Python platform that exposes 1,000+ causal-inference and\neconometrics functions behind one unified API, plus a machine-readable MCP\nserver an agent can drive directly.\n\nIt is **one more choice**, not a replacement. The Stata \u002F R \u002F Python templates in\n`templates\u002F` remain the default for users who want drop-in, version-pinned\nscripts. Reach for StatsPAI when you want a single Python surface that covers the\nwhole AER identification toolkit, self-describes its assumptions to an agent, and\nexports publication-ready LaTeX \u002F Word \u002F Excel tables from the estimator object.\n\nThis skill does **not** override the methodology. The modern-default rules in\n`aer-identification` (no TWFE on staggered data, Anderson-Rubin under weak IV,\nlocal-linear RDD, placebo inference for SCM) still bind. StatsPAI executes those\nrules; it does not relax them.\n\n## When to Use\n\n- You want to **run** the empirical analysis interactively from the agent, not\n  just receive template code to run later by hand\n- You want one Python dependency covering DiD, IV, RDD, SCM, matching, DML, and\n  causal forests instead of stitching together `fixest`, `did`, `rdrobust`,\n  `linearmodels`, and `scdata`\n- A StatsPAI MCP server is connected (tools prefixed `statspai`) and you can\n  chain `detect_design → recommend → fit → audit_result → sensitivity → bibtex`\n- You need publication-ready tables straight from the estimator object\n  (`.to_latex()`, `.to_docx()`, `.to_excel()`) to hand to `aer-tables-figures`\n\n## When NOT to Use\n\n- The deposited replication package must be **Stata-only** (some coauthors \u002F\n  Data Editor workflows assume `.do` files) → use `templates\u002Fstata\u002F`\n- You want exact, pinned, language-native scripts to commit verbatim into an\n  openICPSR deposit → use `templates\u002Fpython\u002F`, `templates\u002Fr\u002F`, or\n  `templates\u002Fstata\u002F`\n- The methodological *choice* is the open question — go to `aer-identification`\n  first; come back here to execute once the design is fixed\n\n## Install and Connect\n\nTwo ways to use StatsPAI; they share the same estimators.\n\n```bash\npip install statspai\n```\n\n```python\nimport statspai as sp\n\ndf = sp.datasets.mpdta()                       # bundled teaching panel\ncs = sp.callaway_santanna(data=df, y=\"lemp\", t=\"year\",\n                          i=\"countyreal\", g=\"first_treat\")\nprint(sp.aggte(cs, type=\"simple\").summary())   # staggered-robust ATT\n```\n\n**Agent-native (MCP) path.** When a StatsPAI MCP server is connected, the agent\ncalls the same methods as tools (`statspai` namespace) and gets back structured\nresult handles instead of parsing console output. This is the preferred path\ninside an agent loop because every function publishes its assumptions,\npreconditions, and failure modes for the agent to read before it runs.\n\n## Recommended Agent-Native Workflow\n\nDrive the MCP server as a chain, not as one-shot calls. Pass `as_handle=true`\n(or `detail='agent'`) so each step returns a `result_id` the next step consumes —\nno need to ferry betas, standard errors, or covariance matrices by hand.\n\n1. **`detect_design`** — infer the study shape (DiD \u002F IV \u002F RDD \u002F SCM \u002F panel), or\n   pass `design=` explicitly when you already know it.\n2. **`preflight` + `recommend`** — surface design problems (no never-treated\n   group, weak first stage, manipulation at the cutoff) and let the engine\n   propose the modern-default estimator.\n3. **Fit with `as_handle=true`** — e.g. `callaway_santanna`, `ivreg`,\n   `rdrobust`, `synth`. You get a `result_id` you can chain downstream.\n4. **`audit_result(result_id=...)`** — enumerate the robustness checks still\n   missing; for each, call the `suggest_function` it emits. This is the\n   referee-anticipation step from `aer-robustness`, automated.\n5. **`honest_did_from_result` \u002F `sensitivity_from_result`** — design-specific\n   sensitivity (Rambachan-Roth honest bounds, Oster δ, etc.) directly off the\n   handle.\n6. **`bibtex(keys=[...])`** — pull verified citations for every estimator and\n   diagnostic you used. **Never invent references** — `paper.bib` is the single\n   source of truth.\n\nToken economy: pass `detail='minimal'` on cheap sub-step calls; the default\n`detail='agent'` carries the violations list and `next_steps` you actually need.\n\n## Worked Execution Snapshot\n\nBefore handing results to `aer-robustness`, leave this shape:\n\n```text\nDESIGN: staggered DiD; n = 48,212; cohorts = 37\nRESULT: ATT = -0.042; SE = 0.011; pretrend p = 0.64\nAUDIT: forbidden weight = 0.31; honest-DiD Mbar=1 bound [-0.071, -0.009]\nDECISION: advance only if template cross-check agrees within 0.002\n```\n\n## Mapping AER Identification Strategies to StatsPAI\n\nEach row keeps the modern default from `aer-identification`; StatsPAI is the\nexecution surface.\n\n| Strategy | Modern default (see `aer-identification`) | StatsPAI entry point |\n|---|---|---|\n| Staggered DiD | Callaway-Sant'Anna ATT(g,t); never raw TWFE | `callaway_santanna`, `aggte`, `did_imputation`, `sun_abraham`, `did_multiplegt`, `lp_did` |\n| Forbidden-comparison check | Goodman-Bacon decomposition | `bacon_decomposition`, `bacon_plot` |\n| Event study \u002F pre-trends | Joint pre-period test, not just the plot | `event_study`, `pretrends_test`, `honest_did` |\n| IV \u002F weak instruments | Anderson-Rubin, not first-stage F > 10 | `ivreg`, `anderson_rubin_ci`, `effective_f_test`, `tF_adjustment` |\n| Shift-share \u002F Bartik | Rotemberg weights or shock-level inference | `bartik` |\n| RDD | Local-linear, MSE-optimal bandwidth, RBC CI | `rdrobust`, `rdbwselect`, `rdplot`, `rddensity` (McCrary) |\n| Synthetic control | Placebo inference; modern variants | `synth`, `gsynth`, `augsynth`, `sdid`, `synth_time_placebo`, `synth_loo` |\n| Factor-model counterfactuals | Interactive FE \u002F matrix completion for larger treated blocks | `interactive_fe`, `matrix_completion` |\n| Bunching \u002F kink designs | Excess mass over a polynomial counterfactual | `bunching`, `notch` |\n| DML \u002F causal ML | Cross-fit nuisance; honest CIs | `dml`, `causal_forest`, `metalearner`, `tmle`, `aipw` |\n| Distributional effects | Quantile treatment effects when the mean hides the action | `qte` |\n| Robustness \u002F few-cluster inference | Specification curve; cluster-robust variance; exact randomization tests | `spec_curve`, `wild_cluster_bootstrap`, `twoway_cluster`, `conley`, `ri_test` |\n| Sensitivity to confounding \u002F multiplicity | Oster δ-R²; Cinelli-Hazlett robustness value; FWER control | `oster_delta`, `oster_bounds`, `robustness_value`, `romano_wolf` |\n| Partial identification under attrition | Lee (2009) trimming bounds when selection is differential | `lee_bounds` |\n\nWhen in doubt about *whether* an estimator is appropriate, that decision belongs\nto `aer-identification`. This table is for *how* to run the one you've chosen.\n\n## Robustness, Heterogeneity, Sensitivity\n\n`audit_result` is the bridge to `aer-robustness`: it reads the fitted handle and\nlists the checks a referee will expect — placebo, alternative samples,\nheterogeneity by cohort, leave-one-out for SCM — emitting a `suggest_function`\nfor each. Drive that list to closure rather than guessing which robustness\nchecks to add.\n\nFor design-specific sensitivity, prefer the `*_from_result` tools so you never\nre-specify the model:\n\n- `honest_did_from_result` — Rambachan-Roth (2023) honest bounds for DiD\n- `sensitivity_from_result` — Oster (2019) δ \u002F unobserved-confounding bounds\n- `robustness_value` — Cinelli-Hazlett (2020) partial-R² robustness value and bias factor\n- `lee_bounds` — Lee (2009) trimming bounds under differential attrition\n- `evalue_from_result` — E-value for observational designs\n\n## Publication Export — Hand Off to `aer-tables-figures`\n\nStatsPAI result objects export directly, which removes manual table assembly:\n\n```python\nres.to_latex(\"output\u002Ftables\u002Ftable3.tex\")   # AER booktabs-compatible\nres.to_docx(\"output\u002Ftables\u002Ftable3.docx\")\nres.to_excel(\"output\u002Ftables\u002Ftable3.xlsx\")\n```\n\nThe **house-style rules still come from `aer-tables-figures`** — column count,\nnote structure, significance-star policy, and the booktabs conventions. Use\nStatsPAI to *emit* the table; use `aer-tables-figures` to decide what the table\nshould look like before you ship it.\n\n## Validation Status — Read It Before Trusting a Number\n\nStatsPAI labels every function with a `validation_status`. Treat these as a\ntrust tier, not decoration:\n\n- **Certified** numerical evidence (benchmarked against Stata \u002F R reference\n  implementations) → safe for a main specification.\n- **API-stable breadth** (broad coverage, not yet numerically certified\n  end-to-end) → fine for exploration; **re-run the headline result against a\n  reference implementation** (`templates\u002F`) before it becomes a main table in a\n  top-5 submission.\n\nFor an AER main result, cross-check at least the headline coefficient and its\ninference against the language-native template. Convergent numbers across two\nengines is itself a robustness signal a referee will respect.\n\n## Red Flags\n\n- Reporting a StatsPAI estimate as a main AER specification without confirming\n  its `validation_status` is certified or cross-checking against `templates\u002F`\n- Letting the unified API tempt you into a *worse* estimator than\n  `aer-identification` prescribes (e.g. plain TWFE because it is one call)\n- Skipping `audit_result` and hand-picking robustness checks — the whole point\n  of the agent-native loop is that it enumerates what you missed\n- Pasting a citation StatsPAI did not return from `bibtex` — verified keys only\n\n## Repository Resources\n\nWhen working from the AER-skills repository or plugin bundle, load only the\nrelevant resource:\n\n- Estimator defaults, diagnostics, and primary citations: `docs\u002Fmethods-reference.md`\n- The methodological decision that precedes execution: `skills\u002Faer-identification\u002FSKILL.md`\n- Referee-anticipating checks `audit_result` should reproduce: `skills\u002Faer-robustness\u002FSKILL.md`\n- House-style rules for the tables StatsPAI emits: `skills\u002Faer-tables-figures\u002FSKILL.md`\n- Reference Python implementation to cross-check headline numbers: `templates\u002Fpython\u002Fmain_did.py`\n- Pinned reference-engine dependencies: `templates\u002Fpython\u002Frequirements.txt`\n\nUse `aer-identification` to fix the design before running anything here; use\n`docs\u002Fmethods-reference.md` to confirm the estimand, diagnostic, and citation\nthe manuscript must report.\n\n## Handoff\n\n```text\nENGINE: StatsPAI (python-api | mcp)\nDESIGN: \u003CDiD | IV | RDD | SCM | DML | ...>\nESTIMATOR FUNCTION: \u003Cstatspai function used>\nVALIDATION STATUS: \u003Ccertified | api-stable — cross-checked? yes\u002Fno>\nAUDIT_RESULT CHECKS CLOSED: \u003Clist or \"none\">\nSENSITIVITY RUN: \u003Chonest_did | oster | evalue | none>\nTABLES EXPORTED: \u003Cpaths or \"none\">\nNEXT SKILL: aer-robustness (close audit) → aer-tables-figures (house style)\n```\n\n## Anti-Patterns\n\n- Treating `aer-statspai` as a substitute for `aer-identification` — it executes\n  the design, it does not choose it\n- Shipping an `api-stable` result as a main top-5 table with no reference\n  cross-check\n- Using the unified API to add a fifth \"robustness\" estimator that shares the\n  same identifying variation as the main result and calling it confirmation\n- Exporting a table from StatsPAI and skipping `aer-tables-figures` — the engine\n  emits a table, it does not enforce AER house style\n","# AER StatsPAI\n\n## Overview\n\n`aer-statspai` is the **implementation engine** option for this stack. Where\n`aer-identification` and `aer-robustness` decide *which* estimator and *which*\ndiagnostics a referee will demand, this skill is about *running* them — through\n[StatsPAI](https:\u002F\u002Fgithub.com\u002Fbrycewang-stanford\u002FStatsPAI), an open-source,\nagent-native Python platform that exposes 1,000+ causal-inference and\neconometrics functions behind one unified API, plus a machine-readable MCP\nserver an agent can drive directly.\n\nIt is **one more choice**, not a replacement. The Stata \u002F R \u002F Python templates in\n`templates\u002F` remain the default for users who want drop-in, version-pinned\nscripts. Reach for StatsPAI when you want a single Python surface that covers the\nwhole AER identification toolkit, self-describes its assumptions to an agent, and\nexports publication-ready LaTeX \u002F Word \u002F Excel tables from the estimator object.\n\nThis skill does **not** override the methodology. The modern-default rules in\n`aer-identification` (no TWFE on staggered data, Anderson-Rubin under weak IV,\nlocal-linear RDD, placebo inference for SCM) still bind. StatsPAI executes those\nrules; it does not relax them.\n\n## When to Use\n\n- You want to **run** the empirical analysis interactively from the agent, not\n  just receive template code to run later by hand\n- You want one Python dependency covering DiD, IV, RDD, SCM, matching, DML, and\n  causal forests instead of stitching together `fixest`, `did`, `rdrobust`,\n  `linearmodels`, and `scdata`\n- A StatsPAI MCP server is connected (tools prefixed `statspai`) and you can\n  chain `detect_design → recommend → fit → audit_result → sensitivity → bibtex`\n- You need publication-ready tables straight from the estimator object\n  (`.to_latex()`, `.to_docx()`, `.to_excel()`) to hand to `aer-tables-figures`\n\n## When NOT to Use\n\n- The deposited replication package must be **Stata-only** (some coauthors \u002F\n  Data Editor workflows assume `.do` files) → use `templates\u002Fstata\u002F`\n- You want exact, pinned, language-native scripts to commit verbatim into an\n  openICPSR deposit → use `templates\u002Fpython\u002F`, `templates\u002Fr\u002F`, or\n  `templates\u002Fstata\u002F`\n- The methodological *choice* is the open question — go to `aer-identification`\n  first; come back here to execute once the design is fixed\n\n## Install and Connect\n\nTwo ways to use StatsPAI; they share the same estimators.\n\n```bash\npip install statspai\n```\n\n```python\nimport statspai as sp\n\ndf = sp.datasets.mpdta()                       # bundled teaching panel\ncs = sp.callaway_santanna(data=df, y=\"lemp\", t=\"year\",\n                          i=\"countyreal\", g=\"first_treat\")\nprint(sp.aggte(cs, type=\"simple\").summary())   # staggered-robust ATT\n```\n\n**Agent-native (MCP) path.** When a StatsPAI MCP server is connected, the agent\ncalls the same methods as tools (`statspai` namespace) and gets back structured\nresult handles instead of parsing console output. This is the preferred path\ninside an agent loop because every function publishes its assumptions,\npreconditions, and failure modes for the agent to read before it runs.\n\n## Recommended Agent-Native Workflow\n\nDrive the MCP server as a chain, not as one-shot calls. Pass `as_handle=true`\n(or `detail='agent'`) so each step returns a `result_id` the next step consumes —\nno need to ferry betas, standard errors, or covariance matrices by hand.\n\n1. **`detect_design`** — infer the study shape (DiD \u002F IV \u002F RDD \u002F SCM \u002F panel), or\n   pass `design=` explicitly when you already know it.\n2. **`preflight` + `recommend`** — surface design problems (no never-treated\n   group, weak first stage, manipulation at the cutoff) and let the engine\n   propose the modern-default estimator.\n3. **Fit with `as_handle=true`** — e.g. `callaway_santanna`, `ivreg`,\n   `rdrobust`, `synth`. You get a `result_id` you can chain downstream.\n4. **`audit_result(result_id=...)`** — enumerate the robustness checks still\n   missing; for each, call the `suggest_function` it emits. This is the\n   referee-anticipation step from `aer-robustness`, automated.\n5. **`honest_did_from_result` \u002F `sensitivity_from_result`** — design-specific\n   sensitivity (Rambachan-Roth honest bounds, Oster δ, etc.) directly off the\n   handle.\n6. **`bibtex(keys=[...])`** — pull verified citations for every estimator and\n   diagnostic you used. **Never invent references** — `paper.bib` is the single\n   source of truth.\n\nToken economy: pass `detail='minimal'` on cheap sub-step calls; the default\n`detail='agent'` carries the violations list and `next_steps` you actually need.\n\n## Worked Execution Snapshot\n\nBefore handing results to `aer-robustness`, leave this shape:\n\n```text\nDESIGN: staggered DiD; n = 48,212; cohorts = 37\nRESULT: ATT = -0.042; SE = 0.011; pretrend p = 0.64\nAUDIT: forbidden weight = 0.31; honest-DiD Mbar=1 bound [-0.071, -0.009]\nDECISION: advance only if template cross-check agrees within 0.002\n```\n\n## Mapping AER Identification Strategies to StatsPAI\n\nEach row keeps the modern default from `aer-identification`; StatsPAI is the\nexecution surface.\n\n| Strategy | Modern default (see `aer-identification`) | StatsPAI entry point |\n|---|---|---|\n| Staggered DiD | Callaway-Sant'Anna ATT(g,t); never raw TWFE | `callaway_santanna`, `aggte`, `did_imputation`, `sun_abraham`, `did_multiplegt`, `lp_did` |\n| Forbidden-comparison check | Goodman-Bacon decomposition | `bacon_decomposition`, `bacon_plot` |\n| Event study \u002F pre-trends | Joint pre-period test, not just the plot | `event_study`, `pretrends_test`, `honest_did` |\n| IV \u002F weak instruments | Anderson-Rubin, not first-stage F > 10 | `ivreg`, `anderson_rubin_ci`, `effective_f_test`, `tF_adjustment` |\n| Shift-share \u002F Bartik | Rotemberg weights or shock-level inference | `bartik` |\n| RDD | Local-linear, MSE-optimal bandwidth, RBC CI | `rdrobust`, `rdbwselect`, `rdplot`, `rddensity` (McCrary) |\n| Synthetic control | Placebo inference; modern variants | `synth`, `gsynth`, `augsynth`, `sdid`, `synth_time_placebo`, `synth_loo` |\n| Factor-model counterfactuals | Interactive FE \u002F matrix completion for larger treated blocks | `interactive_fe`, `matrix_completion` |\n| Bunching \u002F kink designs | Excess mass over a polynomial counterfactual | `bunching`, `notch` |\n| DML \u002F causal ML | Cross-fit nuisance; honest CIs | `dml`, `causal_forest`, `metalearner`, `tmle`, `aipw` |\n| Distributional effects | Quantile treatment effects when the mean hides the action | `qte` |\n| Robustness \u002F few-cluster inference | Specification curve; cluster-robust variance; exact randomization tests | `spec_curve`, `wild_cluster_bootstrap`, `twoway_cluster`, `conley`, `ri_test` |\n| Sensitivity to confounding \u002F multiplicity | Oster δ-R²; Cinelli-Hazlett robustness value; FWER control | `oster_delta`, `oster_bounds`, `robustness_value`, `romano_wolf` |\n| Partial identification under attrition | Lee (2009) trimming bounds when selection is differential | `lee_bounds` |\n\nWhen in doubt about *whether* an estimator is appropriate, that decision belongs\nto `aer-identification`. This table is for *how* to run the one you've chosen.\n\n## Robustness, Heterogeneity, Sensitivity\n\n`audit_result` is the bridge to `aer-robustness`: it reads the fitted handle and\nlists the checks a referee will expect — placebo, alternative samples,\nheterogeneity by cohort, leave-one-out for SCM — emitting a `suggest_function`\nfor each. Drive that list to closure rather than guessing which robustness\nchecks to add.\n\nFor design-specific sensitivity, prefer the `*_from_result` tools so you never\nre-specify the model:\n\n- `honest_did_from_result` — Rambachan-Roth (2023) honest bounds for DiD\n- `sensitivity_from_result` — Oster (2019) δ \u002F unobserved-confounding bounds\n- `robustness_value` — Cinelli-Hazlett (2020) partial-R² robustness value and bias factor\n- `lee_bounds` — Lee (2009) trimming bounds under differential attrition\n- `evalue_from_result` — E-value for observational designs\n\n## Publication Export — Hand Off to `aer-tables-figures`\n\nStatsPAI result objects export directly, which removes manual table assembly:\n\n```python\nres.to_latex(\"output\u002Ftables\u002Ftable3.tex\")   # AER booktabs-compatible\nres.to_docx(\"output\u002Ftables\u002Ftable3.docx\")\nres.to_excel(\"output\u002Ftables\u002Ftable3.xlsx\")\n```\n\nThe **house-style rules still come from `aer-tables-figures`** — column count,\nnote structure, significance-star policy, and the booktabs conventions. Use\nStatsPAI to *emit* the table; use `aer-tables-figures` to decide what the table\nshould look like before you ship it.\n\n## Validation Status — Read It Before Trusting a Number\n\nStatsPAI labels every function with a `validation_status`. Treat these as a\ntrust tier, not decoration:\n\n- **Certified** numerical evidence (benchmarked against Stata \u002F R reference\n  implementations) → safe for a main specification.\n- **API-stable breadth** (broad coverage, not yet numerically certified\n  end-to-end) → fine for exploration; **re-run the headline result against a\n  reference implementation** (`templates\u002F`) before it becomes a main table in a\n  top-5 submission.\n\nFor an AER main result, cross-check at least the headline coefficient and its\ninference against the language-native template. Convergent numbers across two\nengines is itself a robustness signal a referee will respect.\n\n## Red Flags\n\n- Reporting a StatsPAI estimate as a main AER specification without confirming\n  its `validation_status` is certified or cross-checking against `templates\u002F`\n- Letting the unified API tempt you into a *worse* estimator than\n  `aer-identification` prescribes (e.g. plain TWFE because it is one call)\n- Skipping `audit_result` and hand-picking robustness checks — the whole point\n  of the agent-native loop is that it enumerates what you missed\n- Pasting a citation StatsPAI did not return from `bibtex` — verified keys only\n\n## Repository Resources\n\nWhen working from the AER-skills repository or plugin bundle, load only the\nrelevant resource:\n\n- Estimator defaults, diagnostics, and primary citations: `docs\u002Fmethods-reference.md`\n- The methodological decision that precedes execution: `skills\u002Faer-identification\u002FSKILL.md`\n- Referee-anticipating checks `audit_result` should reproduce: `skills\u002Faer-robustness\u002FSKILL.md`\n- House-style rules for the tables StatsPAI emits: `skills\u002Faer-tables-figures\u002FSKILL.md`\n- Reference Python implementation to cross-check headline numbers: `templates\u002Fpython\u002Fmain_did.py`\n- Pinned reference-engine dependencies: `templates\u002Fpython\u002Frequirements.txt`\n\nUse `aer-identification` to fix the design before running anything here; use\n`docs\u002Fmethods-reference.md` to confirm the estimand, diagnostic, and citation\nthe manuscript must report.\n\n## Handoff\n\n```text\nENGINE: StatsPAI (python-api | mcp)\nDESIGN: \u003CDiD | IV | RDD | SCM | DML | ...>\nESTIMATOR FUNCTION: \u003Cstatspai function used>\nVALIDATION STATUS: \u003Ccertified | api-stable — cross-checked? yes\u002Fno>\nAUDIT_RESULT CHECKS CLOSED: \u003Clist or \"none\">\nSENSITIVITY RUN: \u003Chonest_did | oster | evalue | none>\nTABLES EXPORTED: \u003Cpaths or \"none\">\nNEXT SKILL: aer-robustness (close audit) → aer-tables-figures (house style)\n```\n\n## Anti-Patterns\n\n- Treating `aer-statspai` as a substitute for `aer-identification` — it executes\n  the design, it does not choose it\n- Shipping an `api-stable` result as a main top-5 table with no reference\n  cross-check\n- Using the unified API to add a fifth \"robustness\" estimator that shares the\n  same identifying variation as the main result and calling it confirmation\n- Exporting a table from StatsPAI and skipping `aer-tables-figures` — the engine\n  emits a table, it does not enforce AER house style\n",{"name":5,"description":7},"main","18f52829480a4e8df7dd5d9c13ce3b5284311891","skills\u002Fresearch\u002Faer-statspai\u002F1.0.0","https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog","https:\u002F\u002Fgithub.com\u002Fvc999999999","2026-10-11 20:16:46",{"source":8,"skillName":5,"commands":32,"prompt":42,"zipAvailable":43},[33,36,39],{"tool":34,"command":35},"npx","npx skills add vc999999999\u002Fskillflux-catalog --skill aer-statspai",{"tool":37,"command":38},"bunx","bunx skills add vc999999999\u002Fskillflux-catalog --skill aer-statspai",{"tool":40,"command":41},"pnpm","pnpm dlx skills add vc999999999\u002Fskillflux-catalog --skill aer-statspai","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002FSKILL.md",true,[45,51,55,59,63,67,73,77,81,85,89,93,97,101,105,109,113,117,121,127,131,137,141,145,149,153,157,161,163,166,170,174,178,182,186,190,194,198,202,206,210,214,218,222,226,230,234,238,242,246,250,254,258,262,266,270,274,278,282,286,290,294,298,302,306],{"path":46,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":49,"sourceHtmlUrl":50},"agents\u002Fopenai.yaml","REFERENCE","参考文档",228,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fagents\u002Fopenai.yaml",{"path":52,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":53,"sourceHtmlUrl":54},"docs\u002Fdesign-principles.md",4887,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fdocs\u002Fdesign-principles.md",{"path":56,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":57,"sourceHtmlUrl":58},"docs\u002Fglossary.md",20363,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fdocs\u002Fglossary.md",{"path":60,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":61,"sourceHtmlUrl":62},"docs\u002Fmethods-reference.md",16629,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fdocs\u002Fmethods-reference.md",{"path":64,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":65,"sourceHtmlUrl":66},"docs\u002Fsource-register.md",4294,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fdocs\u002Fsource-register.md",{"path":68,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":71,"sourceHtmlUrl":72},"examples\u002F_aer_numeric_check.py","SCRIPT","脚本",2570,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fexamples\u002F_aer_numeric_check.py",{"path":74,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":75,"sourceHtmlUrl":76},"examples\u002Faer-exemplars.md",8845,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fexamples\u002Faer-exemplars.md",{"path":78,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":79,"sourceHtmlUrl":80},"examples\u002Fdml-plr-demo\u002Fdml_plr_demo.py",9775,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fexamples\u002Fdml-plr-demo\u002Fdml_plr_demo.py",{"path":82,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":83,"sourceHtmlUrl":84},"examples\u002Fdml-plr-demo\u002FREADME.md",2804,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fexamples\u002Fdml-plr-demo\u002FREADME.md",{"path":86,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":87,"sourceHtmlUrl":88},"examples\u002Flee-bounds-demo\u002Flee_bounds_demo.py",9892,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fexamples\u002Flee-bounds-demo\u002Flee_bounds_demo.py",{"path":90,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":91,"sourceHtmlUrl":92},"examples\u002Flee-bounds-demo\u002FREADME.md",3557,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fexamples\u002Flee-bounds-demo\u002FREADME.md",{"path":94,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":95,"sourceHtmlUrl":96},"examples\u002Flp-did-demo\u002Flp_did_demo.py",13150,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fexamples\u002Flp-did-demo\u002Flp_did_demo.py",{"path":98,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":99,"sourceHtmlUrl":100},"examples\u002Flp-did-demo\u002FREADME.md",2996,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fexamples\u002Flp-did-demo\u002FREADME.md",{"path":102,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":103,"sourceHtmlUrl":104},"examples\u002Fmodern-aer-exemplars.md",18601,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fexamples\u002Fmodern-aer-exemplars.md",{"path":106,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":107,"sourceHtmlUrl":108},"examples\u002Foster-ovb-demo\u002Foster_ovb_demo.py",7451,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fexamples\u002Foster-ovb-demo\u002Foster_ovb_demo.py",{"path":110,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":111,"sourceHtmlUrl":112},"examples\u002Foster-ovb-demo\u002FREADME.md",2131,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fexamples\u002Foster-ovb-demo\u002FREADME.md",{"path":114,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":115,"sourceHtmlUrl":116},"examples\u002Fpower-mde-demo\u002Fpower_mde_demo.py",7894,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fexamples\u002Fpower-mde-demo\u002Fpower_mde_demo.py",{"path":118,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":119,"sourceHtmlUrl":120},"examples\u002Fpower-mde-demo\u002FREADME.md",3159,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fexamples\u002Fpower-mde-demo\u002FREADME.md",{"path":122,"resourceType":123,"resourceTypeDesc":124,"sizeBytes":125,"sourceHtmlUrl":126},"examples\u002Freplication-package-skeleton\u002FLICENSE","LICENSE","许可证",962,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fexamples\u002Freplication-package-skeleton\u002FLICENSE",{"path":128,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":129,"sourceHtmlUrl":130},"examples\u002Freplication-package-skeleton\u002FREADME.md",6913,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fexamples\u002Freplication-package-skeleton\u002FREADME.md",{"path":132,"resourceType":133,"resourceTypeDesc":134,"sizeBytes":135,"sourceHtmlUrl":136},"examples\u002Freplication-package-skeleton\u002Frun_all.do","OTHER","其他",620,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fexamples\u002Freplication-package-skeleton\u002Frun_all.do",{"path":138,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":139,"sourceHtmlUrl":140},"examples\u002Fsensitivity-rv-demo\u002FREADME.md",3693,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fexamples\u002Fsensitivity-rv-demo\u002FREADME.md",{"path":142,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":143,"sourceHtmlUrl":144},"examples\u002Fsensitivity-rv-demo\u002Fsensitivity_rv_demo.py",11475,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fexamples\u002Fsensitivity-rv-demo\u002Fsensitivity_rv_demo.py",{"path":146,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":147,"sourceHtmlUrl":148},"examples\u002Fstaggered-did-demo\u002FREADME.md",1464,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fexamples\u002Fstaggered-did-demo\u002FREADME.md",{"path":150,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":151,"sourceHtmlUrl":152},"examples\u002Fstaggered-did-demo\u002Fstaggered_did_demo.py",9303,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fexamples\u002Fstaggered-did-demo\u002Fstaggered_did_demo.py",{"path":154,"resourceType":133,"resourceTypeDesc":134,"sizeBytes":155,"sourceHtmlUrl":156},"examples\u002Fstaggered-did-demo\u002Fstaggered_did_demo.R",6882,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fexamples\u002Fstaggered-did-demo\u002Fstaggered_did_demo.R",{"path":158,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":159,"sourceHtmlUrl":160},"examples\u002Fsun-abraham-demo\u002FREADME.md",3778,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fexamples\u002Fsun-abraham-demo\u002FREADME.md",{"path":123,"resourceType":123,"resourceTypeDesc":124,"sizeBytes":162,"sourceHtmlUrl":16},1067,{"path":164,"resourceType":123,"resourceTypeDesc":124,"sizeBytes":162,"sourceHtmlUrl":165},"LICENSE.txt","https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002FLICENSE.txt",{"path":167,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":168,"sourceHtmlUrl":169},"scripts\u002Fscaffold_project.py",4070,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fscripts\u002Fscaffold_project.py",{"path":171,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":172,"sourceHtmlUrl":173},"scripts\u002Fstatspai_tools.txt",2116,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fscripts\u002Fstatspai_tools.txt",{"path":175,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":176,"sourceHtmlUrl":177},"skillflux.json",1162,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fskillflux.json",{"path":179,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":180,"sourceHtmlUrl":181},"skillflux.review.json",2386,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Fskillflux.review.json",{"path":183,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":184,"sourceHtmlUrl":185},"templates\u002Fpython\u002Fclean.py",783,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fpython\u002Fclean.py",{"path":187,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":188,"sourceHtmlUrl":189},"templates\u002Fpython\u002Fdescriptives.py",674,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fpython\u002Fdescriptives.py",{"path":191,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":192,"sourceHtmlUrl":193},"templates\u002Fpython\u002Ffigures.py",236,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fpython\u002Ffigures.py",{"path":195,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":196,"sourceHtmlUrl":197},"templates\u002Fpython\u002Fheterogeneity.py",249,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fpython\u002Fheterogeneity.py",{"path":199,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":200,"sourceHtmlUrl":201},"templates\u002Fpython\u002Fmain_did.py",5061,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fpython\u002Fmain_did.py",{"path":203,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":204,"sourceHtmlUrl":205},"templates\u002Fpython\u002FREADME.md",2527,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fpython\u002FREADME.md",{"path":207,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":208,"sourceHtmlUrl":209},"templates\u002Fpython\u002Frequirements.txt",876,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fpython\u002Frequirements.txt",{"path":211,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":212,"sourceHtmlUrl":213},"templates\u002Fpython\u002Frobustness.py",267,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fpython\u002Frobustness.py",{"path":215,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":216,"sourceHtmlUrl":217},"templates\u002Fpython\u002Frun_all.py",2097,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fpython\u002Frun_all.py",{"path":219,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":220,"sourceHtmlUrl":221},"templates\u002Fpython\u002Fsetup.py",1438,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fpython\u002Fsetup.py",{"path":223,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":224,"sourceHtmlUrl":225},"templates\u002Fpython\u002Ftables.py",1909,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fpython\u002Ftables.py",{"path":227,"resourceType":133,"resourceTypeDesc":134,"sizeBytes":228,"sourceHtmlUrl":229},"templates\u002Fr\u002F00_setup.R",2356,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fr\u002F00_setup.R",{"path":231,"resourceType":133,"resourceTypeDesc":134,"sizeBytes":232,"sourceHtmlUrl":233},"templates\u002Fr\u002F01_clean.R",766,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fr\u002F01_clean.R",{"path":235,"resourceType":133,"resourceTypeDesc":134,"sizeBytes":236,"sourceHtmlUrl":237},"templates\u002Fr\u002F02_descriptives.R",682,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fr\u002F02_descriptives.R",{"path":239,"resourceType":133,"resourceTypeDesc":134,"sizeBytes":240,"sourceHtmlUrl":241},"templates\u002Fr\u002F03_main_did.R",6105,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fr\u002F03_main_did.R",{"path":243,"resourceType":133,"resourceTypeDesc":134,"sizeBytes":244,"sourceHtmlUrl":245},"templates\u002Fr\u002F04_robustness.R",279,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fr\u002F04_robustness.R",{"path":247,"resourceType":133,"resourceTypeDesc":134,"sizeBytes":248,"sourceHtmlUrl":249},"templates\u002Fr\u002F05_heterogeneity.R",261,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fr\u002F05_heterogeneity.R",{"path":251,"resourceType":133,"resourceTypeDesc":134,"sizeBytes":252,"sourceHtmlUrl":253},"templates\u002Fr\u002F06_tables.R",1987,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fr\u002F06_tables.R",{"path":255,"resourceType":133,"resourceTypeDesc":134,"sizeBytes":256,"sourceHtmlUrl":257},"templates\u002Fr\u002F07_figures.R",248,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fr\u002F07_figures.R",{"path":259,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":260,"sourceHtmlUrl":261},"templates\u002Fr\u002FREADME.md",1942,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fr\u002FREADME.md",{"path":263,"resourceType":133,"resourceTypeDesc":134,"sizeBytes":264,"sourceHtmlUrl":265},"templates\u002Fr\u002Frun_all.R",1743,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fr\u002Frun_all.R",{"path":267,"resourceType":133,"resourceTypeDesc":134,"sizeBytes":268,"sourceHtmlUrl":269},"templates\u002Fstata\u002F00_globals.do",1382,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fstata\u002F00_globals.do",{"path":271,"resourceType":133,"resourceTypeDesc":134,"sizeBytes":272,"sourceHtmlUrl":273},"templates\u002Fstata\u002F00_install_packages.do",1092,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fstata\u002F00_install_packages.do",{"path":275,"resourceType":133,"resourceTypeDesc":134,"sizeBytes":276,"sourceHtmlUrl":277},"templates\u002Fstata\u002F01_clean.do",796,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fstata\u002F01_clean.do",{"path":279,"resourceType":133,"resourceTypeDesc":134,"sizeBytes":280,"sourceHtmlUrl":281},"templates\u002Fstata\u002F02_descriptives.do",406,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fstata\u002F02_descriptives.do",{"path":283,"resourceType":133,"resourceTypeDesc":134,"sizeBytes":284,"sourceHtmlUrl":285},"templates\u002Fstata\u002F03_main_did.do",3490,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fstata\u002F03_main_did.do",{"path":287,"resourceType":133,"resourceTypeDesc":134,"sizeBytes":288,"sourceHtmlUrl":289},"templates\u002Fstata\u002F04_robustness.do",314,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fstata\u002F04_robustness.do",{"path":291,"resourceType":133,"resourceTypeDesc":134,"sizeBytes":292,"sourceHtmlUrl":293},"templates\u002Fstata\u002F05_heterogeneity.do",296,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fstata\u002F05_heterogeneity.do",{"path":295,"resourceType":133,"resourceTypeDesc":134,"sizeBytes":296,"sourceHtmlUrl":297},"templates\u002Fstata\u002F06_tables.do",2143,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fstata\u002F06_tables.do",{"path":299,"resourceType":133,"resourceTypeDesc":134,"sizeBytes":300,"sourceHtmlUrl":301},"templates\u002Fstata\u002F07_figures.do",283,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fstata\u002F07_figures.do",{"path":303,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":304,"sourceHtmlUrl":305},"templates\u002Fstata\u002FREADME.md",2055,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fstata\u002FREADME.md",{"path":307,"resourceType":133,"resourceTypeDesc":134,"sizeBytes":308,"sourceHtmlUrl":309},"templates\u002Fstata\u002Frun_all.do",1564,"https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-statspai\u002F1.0.0\u002Ftemplates\u002Fstata\u002Frun_all.do",[311,319,327,335,343,351],{"skillNo":312,"name":313,"nameZh":6,"description":314,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":315,"sourceHtmlUrl":316,"categories":317,"keywords":318,"publishedAt":20},"01M7DKTKB8RM9NFNA9VD5QNM59","scientific-schematics","Generates scientific diagram drafts using Nano Banana 2 AI with smart iterative refinement. Uses Gemini 3.7 Flash for quality review. Refines when the review requests improvement, with at most two generations. Specialized in neural network architectures, system diagrams, flowcharts, biological pathways, and complex scientific visualizations.","https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Fscientific-schematics\u002F1.0.0\u002FLICENSE.md","https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Fscientific-schematics\u002F1.0.0\u002FSKILL.md",[],[],{"skillNo":320,"name":321,"nameZh":6,"description":322,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":323,"sourceHtmlUrl":324,"categories":325,"keywords":326,"publishedAt":20},"01M7DKTE1BMWDNRQ6JHSB8YZVM","scientific-critical-thinking","Evaluates scientific claims and evidence quality. Applies to experimental design validity, biases and confounders, statistical interpretation, evidence grading frameworks (GRADE, Cochrane Risk of Bias), and teaching critical analysis. Supports evidence appraisal and identifying flaws; formal peer review writing belongs to peer-review.","https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Fscientific-critical-thinking\u002F1.0.0\u002FLICENSE.md","https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Fscientific-critical-thinking\u002F1.0.0\u002FSKILL.md",[],[],{"skillNo":328,"name":329,"nameZh":6,"description":330,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":331,"sourceHtmlUrl":332,"categories":333,"keywords":334,"publishedAt":20},"01M7DKT74MD9SFF7K4ZGQJ5PAY","scientific-brainstorming","Facilitates evidence-aware scientific ideation with independent generation, structured discussion, explicit assumptions, transparent evaluation, adversarial review, and decision logs. Use for early-stage research brainstorming or prioritizing candidate directions; hand off empirical validation, study design, ethics or regulatory review, and clinical questions to appropriate experts or skills.","https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Fscientific-brainstorming\u002F1.0.0\u002FLICENSE.md","https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Fscientific-brainstorming\u002F1.0.0\u002FSKILL.md",[],[],{"skillNo":336,"name":337,"nameZh":6,"description":338,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":339,"sourceHtmlUrl":340,"categories":341,"keywords":342,"publishedAt":20},"01M7DKT1837E2PSV5Y6EQV4A4B","aer-tables-figures","Use when constructing or revising regression tables, descriptive statistics tables, or figures after results are estimated and before submission for an AER, AER:Insights, or AEJ manuscript. Implements AER booktabs house style, regression-table layout, and figure-note conventions.","https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-tables-figures\u002F1.0.0\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-tables-figures\u002F1.0.0\u002FSKILL.md",[],[],{"skillNo":344,"name":345,"nameZh":6,"description":346,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":347,"sourceHtmlUrl":348,"categories":349,"keywords":350,"publishedAt":20},"01M7DKSQKV1429VSPVVHAM1QQM","aer-robustness","Use when the main empirical results exist but the manuscript lacks the robustness, heterogeneity, mechanism, and placebo checks that AER referees will demand. Apply after aer-identification and before aer-introduction so that the value-added paragraph can reference these tests.","https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-robustness\u002F1.0.0\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-robustness\u002F1.0.0\u002FSKILL.md",[],[],{"skillNo":352,"name":353,"nameZh":6,"description":354,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":355,"sourceHtmlUrl":356,"categories":357,"keywords":358,"publishedAt":20},"01M7DKSJ0XAZAZCFPRNA0SGDZ1","aer-replication","Use when assembling the AEA Data and Code Availability deposit for an AER, AER:Insights, or AEJ acceptance, writing the README, or auditing a replication package before the AEA Data Editor review. Implements the current AEA policy, including the February 2026 Data and Code Availability Policy.","https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-replication\u002F1.0.0\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fvc999999999\u002Fskillflux-catalog\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002Faer-replication\u002F1.0.0\u002FSKILL.md",[],[]]