[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M79277SGMFEF99FDDAH0XRK1":3,"skill-same:01M79277SGMFEF99FDDAH0XRK1":48},{"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":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":24,"commitSha":25,"skillDirectory":26,"forksCount":27,"repositoryHtmlUrl":28,"repositoryHomepageUrl":29,"ownerHtmlUrl":30,"ownerBlogUrl":31,"ownerDescription":32,"updatedAt":33,"install":34,"resources":47},"01M79277SGMFEF99FDDAH0XRK1","exploratory-data-analysis",null,"Profile and interrogate data before anyone makes a claim about it — a cleaned frame, a folder of parquet parts, a warehouse table you query in place, or an index frame over a corpus of documents. Summary statistics do not identify a distribution — Anscombe's quartet shares a mean, variance and correlation across four unrelated shapes — so nothing is reported that has not been plotted. Covers dtype and cardinality profiling, missingness structure, Spearman against Pearson, a correlation matrix read as blocks instead of skimmed for its biggest cells, subgroup checks and whether one named group, batch or run is an outlier against the rest, variation along whatever index orders the data, writing one row per finding to run\u002Ffindings.jsonl and handing it on to modeling. Use when asked to explore or profile a dataset, before a model is fit, when one group, device, batch or period looks wrong and you need to say how it differs, or when a pile of numbers has to become defensible findings.","pymc-labs\u002Fdaimon","pymc-labs","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F69535651?v=4","01M7925X8T3KFBJ7FJ7V31YCC3","PyMC Labs","01M7925X9W5S549AGZZ126G8JA",24,"MIT","https:\u002F\u002Fgithub.com\u002Fpymc-labs\u002Fdaimon\u002Fblob\u002Fmain\u002Fdefaults\u002Fskills\u002Fexploratory-data-analysis\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: exploratory-data-analysis\ndescription: Profile and interrogate data before anyone makes a claim about it — a cleaned frame, a folder of parquet parts, a warehouse table you query in place, or an index frame over a corpus of documents. Summary statistics do not identify a distribution — Anscombe's quartet shares a mean, variance and correlation across four unrelated shapes — so nothing is reported that has not been plotted. Covers dtype and cardinality profiling, missingness structure, Spearman against Pearson, a correlation matrix read as blocks instead of skimmed for its biggest cells, subgroup checks and whether one named group, batch or run is an outlier against the rest, variation along whatever index orders the data, writing one row per finding to run\u002Ffindings.jsonl and handing it on to modeling. Use when asked to explore or profile a dataset, before a model is fit, when one group, device, batch or period looks wrong and you need to say how it differs, or when a pile of numbers has to become defensible findings.\n---\n\n# Exploratory data analysis\n\nAnscombe's quartet is four datasets sharing a mean, variance, correlation and\nregression line, and four completely different shapes. A statistic you have not\nplotted is not evidence. You look at every column you mean to talk about, and\nappend one row per finding to `run\u002Ffindings.jsonl`.\n\n**Applies to** whatever earlier stages left you: one cleaned frame, parquet\nparts too big to load, a warehouse table you query in place, an index frame over\n500 emails. **Your index may not be a calendar** — trial number, wavelength and\nsequence position order data too, so bucket along the manifest's `grain`.\n**Enter here** once the data is admissible, **stop here** when a description\nanswers it — *why* and *what drives* want a fit, so hand that on to modeling.\n\n## Look at a column before you summarise it\n\n**Never report a mean, a correlation or a slope you have not seen plotted.** A\nsecond mode, a floor at zero, a ceiling at 100, a spike at a `-999` sentinel and\none 10,000× outlier all leave the mean printable and change what it means.\nResolve your input: `run\u002Fclean.parquet`, else `run\u002Fclean\u002F`, else `run\u002Fclean.sql`\n— aggregate with it, plot the `run\u002Fclean_sample.parquet` beside it — else the\nmanifest's source with the dtypes it recorded, else the paste itself.\n\n```python\nimport pandas as pd\ndf = pd.read_parquet(\"run\u002Fclean.parquet\")  # parts or in place: duckdb aggregates, then a sample\nprint(len(df), \"rows\")\nprint(pd.DataFrame({\"dtype\": df.dtypes.astype(str), \"pct_missing\": df.isna().mean().round(3),\n                    \"n_unique\": df.nunique(dropna=True)}))\nprint(df.select_dtypes(\"number\").quantile([0, .01, .25, .5, .75, .99, 1]).T)\n```\n\nPercentiles .01 and .99 beside min and max are where a sentinel or a misplaced\ndecimal shows; `.describe()` hides both. Styling is `pymc-artifact-style`'s.\n\n## Match the method to the question\n\n| Question | Compute | Chart form |\n|---|---|---|\n| What is in this column? | quantiles at 0\u002F.01\u002F.25\u002F.5\u002F.75\u002F.99\u002F1, `nunique`, `skew()` | histogram, 50+ bins; ECDF when the tail is long |\n| What is in this collection? | the index frame's own profile — `n_chars` per document, pixel dims, coordinate bounds, node degree — then quantiles as above | histogram of item size; iterate every item in code, read only a sample into context (`file-handling`); plot coordinates in projected metres, since a degree of longitude shrinks with latitude and there is no basemap here |\n| Do two numerics move together? | `corr(\"pearson\")` and `corr(\"spearman\")` | scatter, with binned medians over it |\n| Do groups differ? | `groupby().agg(n=…, median=…)`, then the same aggregate within each level of anything that could have produced the difference | small multiples or box plots, never a bare bar of means |\n| Does it vary along the index? | count *and* metric per bucket of whatever the `grain` orders — month, trial, wavelength bin, store | line of the row count first, the metric second; drop or label a partial final bucket |\n| Which rows — or which named group — are extreme? | distance from the median in MADs; for a group, summarise per group first, then MADs across the group summaries, and say what `n` groups supports | scatter with the flagged points marked; group summaries as a strip plot |\n\n## Missingness is a variable, not a gap\n\n**Test whether missingness depends on something before you treat it as noise.**\nMissing-at-random and missing-because-of-something print the same null rate and\nmean opposite things.\n\n```python\nmissing = df[\"email\"].isna().rename(\"email_missing\")\nbucket = df[\"created_at\"].dt.to_period(\"M\")   # or trial \u002F\u002F 100, or a wavelength bin\nprint(missing.groupby(df[\"region\"]).mean().round(3))  # by segment\nprint(missing.groupby(bucket).mean().round(3))        # along the index\nprint(df.groupby(missing)[\"amount\"].median())         # vs the outcome\n```\n\nFlat across segments and buckets is collection noise. A jump at one boundary is\na schema change, and that window may not be comparable. A rate that tracks the\noutcome biases any analysis that drops those rows — report it, do not treat it.\n\n## Check every relationship against rank and against subgroup\n\n**Compute Spearman alongside Pearson and read the pairs where they disagree.** A\nlarge gap means the relationship is curved, or that a few points carry the fit.\n\n```python\nnum = df.select_dtypes(\"number\").drop(columns=[\"order_id\"])\npairs = pd.DataFrame({\"pearson\": num.corr(\"pearson\").stack(),\n                      \"spearman\": num.corr(\"spearman\").stack()})\npairs = pairs[pairs.index.get_level_values(0) \u003C pairs.index.get_level_values(1)]\npairs[\"gap\"] = (pairs[\"spearman\"] - pairs[\"pearson\"]).abs()\nprint(pairs.sort_values(\"gap\", ascending=False).round(3).head(10))\nprint(df.groupby(\"region\")[[\"amount\", \"tenure_days\"]].corr(\"spearman\")\n        .xs(\"amount\", level=1)[\"tenure_days\"].round(3))   # a sign no group repeats is noise\n```\n\n## What you must not compute\n\n- **No correlation matrix skimmed for its biggest cells** — 40 columns give 780\n  pairs and the largest are what noise produces. Read it as structure: cluster\n  or factor it, show the reordered heatmap, say how many blocks you kept.\n- **No p-value or significance claim from a comparison you chose after seeing\n  the data.** Report effect size, direction and `n`. A test, a fit or a causal\n  claim is modeling output — modeling is a real next stage when `n` can support\n  one; see the handoff below.\n- **No edits to the frame.** Flag an outlier; `data-cleaning` treats and logs it.\n- **No pattern reported before you ask whether collection made it** — a cliff at\n  a round number, a mode at the form default, a gap on weekends.\n\n## What you write: run\u002Ffindings.jsonl\n\nOne row per finding, appended as you go; `how` is the expression that re-derives\nthe number — pandas, SQL or a shell one-liner. When `n` cannot carry a claim,\nthe finding is the plot, the direction and the size, and the `caveat` says so.\n\n```python\nimport json, os\n\ndef record_finding(id, claim, value, n, how, caveat=None):\n    \"\"\"The ONLY writer for run\u002Ffindings.jsonl. Copy it verbatim; these six keys are\n    the contract. `id` is what eda-storytelling cites a claim by, and `how` is what\n    lets someone re-derive the number — a row named {\"finding\": ...} instead has\n    neither, and the citation check rejects it.\"\"\"\n    os.makedirs(\"run\", exist_ok=True)\n    row = {\"id\": id, \"claim\": claim, \"value\": value, \"n\": int(n), \"how\": how,\n           \"caveat\": caveat}\n    assert set(row) == {\"id\", \"claim\", \"value\", \"n\", \"how\", \"caveat\"}\n    with open(\"run\u002Ffindings.jsonl\", \"a\") as fh:\n        fh.write(json.dumps(row) + \"\\n\")\n\nrecord_finding(\"missing-email-rate\", \"40% of rows have no email\", 0.404, 12043,\n               \"df['email'].isna().mean()\", \"concentrated in the east region\")\n```\n\n`id` is a stable kebab-case slug you choose; `value` may be a number, a string or\na small dict when one number cannot carry the claim. Everything else is required\neven when it feels obvious — a finding with no `n` cannot be rounded honestly,\nand one with no `how` cannot be checked.\n\nCharts stay under `run\u002F`; a notebook to poke at is `marimo_notebooks`'s. You may\nbe the whole answer or one step in it: run this stage, skip what nobody asked\nfor, and if the answer ships from here end the reply with the literal\n`Stages: … · skipped: …` line `eda-storytelling` defines. Entering here on an\nexport someone called clean does not make validation run: it is `skipped` unless\nyou wrote `run\u002Fvalidation.json` this turn.\n\nHand `run\u002Ffindings.jsonl` to `eda-storytelling`, which picks what the reader\nsees, a posterior included — and to modeling when a fit was asked: a\n`pymc-marketing-*` skill (mmm, clv, priors) if one is attached, otherwise the\nsimplest PyMC model that answers the question at this `n`. Name its likelihood\nand priors, say what a real MMM or CLV has that yours lacks (adstock,\nsaturation, a hierarchy), and do not ship it under that name. Record what a\nprofile lacks: the\ngrain one row represents, the index column and its regularity, which columns\ngroup rows, their level counts, what nests in what, each item or covariate's\nunit, missing rate and range, the outcome's distribution and any floor or\nceiling, and every Pearson\u002FSpearman gap that says a relationship is curved.\n\n## Common failure modes\n\n| Symptom | Likely cause | Fix |\n|---|---|---|\n| A \"strong correlation\" that no segment reproduces | pooled Pearson driven by group sizes or a few points | Spearman plus the per-group check above; report the subgroup result |\n| Mean and median disagree wildly, the write-up quotes the mean | heavy tail or an un-flagged outlier | Quote the median with an IQR, flag the tail as its own finding |\n| The reply is 400 lines of `.describe()` and a pairplot | volume substituted for judgement | Cut to findings that change a decision; the rest stays in `run\u002F` |\n| A claim has no `findings.jsonl` row | the number was computed in passing, never recorded | Record it or drop it — `eda-storytelling` rejects unsourced numbers |\n","# Exploratory data analysis\n\nAnscombe's quartet is four datasets sharing a mean, variance, correlation and\nregression line, and four completely different shapes. A statistic you have not\nplotted is not evidence. You look at every column you mean to talk about, and\nappend one row per finding to `run\u002Ffindings.jsonl`.\n\n**Applies to** whatever earlier stages left you: one cleaned frame, parquet\nparts too big to load, a warehouse table you query in place, an index frame over\n500 emails. **Your index may not be a calendar** — trial number, wavelength and\nsequence position order data too, so bucket along the manifest's `grain`.\n**Enter here** once the data is admissible, **stop here** when a description\nanswers it — *why* and *what drives* want a fit, so hand that on to modeling.\n\n## Look at a column before you summarise it\n\n**Never report a mean, a correlation or a slope you have not seen plotted.** A\nsecond mode, a floor at zero, a ceiling at 100, a spike at a `-999` sentinel and\none 10,000× outlier all leave the mean printable and change what it means.\nResolve your input: `run\u002Fclean.parquet`, else `run\u002Fclean\u002F`, else `run\u002Fclean.sql`\n— aggregate with it, plot the `run\u002Fclean_sample.parquet` beside it — else the\nmanifest's source with the dtypes it recorded, else the paste itself.\n\n```python\nimport pandas as pd\ndf = pd.read_parquet(\"run\u002Fclean.parquet\")  # parts or in place: duckdb aggregates, then a sample\nprint(len(df), \"rows\")\nprint(pd.DataFrame({\"dtype\": df.dtypes.astype(str), \"pct_missing\": df.isna().mean().round(3),\n                    \"n_unique\": df.nunique(dropna=True)}))\nprint(df.select_dtypes(\"number\").quantile([0, .01, .25, .5, .75, .99, 1]).T)\n```\n\nPercentiles .01 and .99 beside min and max are where a sentinel or a misplaced\ndecimal shows; `.describe()` hides both. Styling is `pymc-artifact-style`'s.\n\n## Match the method to the question\n\n| Question | Compute | Chart form |\n|---|---|---|\n| What is in this column? | quantiles at 0\u002F.01\u002F.25\u002F.5\u002F.75\u002F.99\u002F1, `nunique`, `skew()` | histogram, 50+ bins; ECDF when the tail is long |\n| What is in this collection? | the index frame's own profile — `n_chars` per document, pixel dims, coordinate bounds, node degree — then quantiles as above | histogram of item size; iterate every item in code, read only a sample into context (`file-handling`); plot coordinates in projected metres, since a degree of longitude shrinks with latitude and there is no basemap here |\n| Do two numerics move together? | `corr(\"pearson\")` and `corr(\"spearman\")` | scatter, with binned medians over it |\n| Do groups differ? | `groupby().agg(n=…, median=…)`, then the same aggregate within each level of anything that could have produced the difference | small multiples or box plots, never a bare bar of means |\n| Does it vary along the index? | count *and* metric per bucket of whatever the `grain` orders — month, trial, wavelength bin, store | line of the row count first, the metric second; drop or label a partial final bucket |\n| Which rows — or which named group — are extreme? | distance from the median in MADs; for a group, summarise per group first, then MADs across the group summaries, and say what `n` groups supports | scatter with the flagged points marked; group summaries as a strip plot |\n\n## Missingness is a variable, not a gap\n\n**Test whether missingness depends on something before you treat it as noise.**\nMissing-at-random and missing-because-of-something print the same null rate and\nmean opposite things.\n\n```python\nmissing = df[\"email\"].isna().rename(\"email_missing\")\nbucket = df[\"created_at\"].dt.to_period(\"M\")   # or trial \u002F\u002F 100, or a wavelength bin\nprint(missing.groupby(df[\"region\"]).mean().round(3))  # by segment\nprint(missing.groupby(bucket).mean().round(3))        # along the index\nprint(df.groupby(missing)[\"amount\"].median())         # vs the outcome\n```\n\nFlat across segments and buckets is collection noise. A jump at one boundary is\na schema change, and that window may not be comparable. A rate that tracks the\noutcome biases any analysis that drops those rows — report it, do not treat it.\n\n## Check every relationship against rank and against subgroup\n\n**Compute Spearman alongside Pearson and read the pairs where they disagree.** A\nlarge gap means the relationship is curved, or that a few points carry the fit.\n\n```python\nnum = df.select_dtypes(\"number\").drop(columns=[\"order_id\"])\npairs = pd.DataFrame({\"pearson\": num.corr(\"pearson\").stack(),\n                      \"spearman\": num.corr(\"spearman\").stack()})\npairs = pairs[pairs.index.get_level_values(0) \u003C pairs.index.get_level_values(1)]\npairs[\"gap\"] = (pairs[\"spearman\"] - pairs[\"pearson\"]).abs()\nprint(pairs.sort_values(\"gap\", ascending=False).round(3).head(10))\nprint(df.groupby(\"region\")[[\"amount\", \"tenure_days\"]].corr(\"spearman\")\n        .xs(\"amount\", level=1)[\"tenure_days\"].round(3))   # a sign no group repeats is noise\n```\n\n## What you must not compute\n\n- **No correlation matrix skimmed for its biggest cells** — 40 columns give 780\n  pairs and the largest are what noise produces. Read it as structure: cluster\n  or factor it, show the reordered heatmap, say how many blocks you kept.\n- **No p-value or significance claim from a comparison you chose after seeing\n  the data.** Report effect size, direction and `n`. A test, a fit or a causal\n  claim is modeling output — modeling is a real next stage when `n` can support\n  one; see the handoff below.\n- **No edits to the frame.** Flag an outlier; `data-cleaning` treats and logs it.\n- **No pattern reported before you ask whether collection made it** — a cliff at\n  a round number, a mode at the form default, a gap on weekends.\n\n## What you write: run\u002Ffindings.jsonl\n\nOne row per finding, appended as you go; `how` is the expression that re-derives\nthe number — pandas, SQL or a shell one-liner. When `n` cannot carry a claim,\nthe finding is the plot, the direction and the size, and the `caveat` says so.\n\n```python\nimport json, os\n\ndef record_finding(id, claim, value, n, how, caveat=None):\n    \"\"\"The ONLY writer for run\u002Ffindings.jsonl. Copy it verbatim; these six keys are\n    the contract. `id` is what eda-storytelling cites a claim by, and `how` is what\n    lets someone re-derive the number — a row named {\"finding\": ...} instead has\n    neither, and the citation check rejects it.\"\"\"\n    os.makedirs(\"run\", exist_ok=True)\n    row = {\"id\": id, \"claim\": claim, \"value\": value, \"n\": int(n), \"how\": how,\n           \"caveat\": caveat}\n    assert set(row) == {\"id\", \"claim\", \"value\", \"n\", \"how\", \"caveat\"}\n    with open(\"run\u002Ffindings.jsonl\", \"a\") as fh:\n        fh.write(json.dumps(row) + \"\\n\")\n\nrecord_finding(\"missing-email-rate\", \"40% of rows have no email\", 0.404, 12043,\n               \"df['email'].isna().mean()\", \"concentrated in the east region\")\n```\n\n`id` is a stable kebab-case slug you choose; `value` may be a number, a string or\na small dict when one number cannot carry the claim. Everything else is required\neven when it feels obvious — a finding with no `n` cannot be rounded honestly,\nand one with no `how` cannot be checked.\n\nCharts stay under `run\u002F`; a notebook to poke at is `marimo_notebooks`'s. You may\nbe the whole answer or one step in it: run this stage, skip what nobody asked\nfor, and if the answer ships from here end the reply with the literal\n`Stages: … · skipped: …` line `eda-storytelling` defines. Entering here on an\nexport someone called clean does not make validation run: it is `skipped` unless\nyou wrote `run\u002Fvalidation.json` this turn.\n\nHand `run\u002Ffindings.jsonl` to `eda-storytelling`, which picks what the reader\nsees, a posterior included — and to modeling when a fit was asked: a\n`pymc-marketing-*` skill (mmm, clv, priors) if one is attached, otherwise the\nsimplest PyMC model that answers the question at this `n`. Name its likelihood\nand priors, say what a real MMM or CLV has that yours lacks (adstock,\nsaturation, a hierarchy), and do not ship it under that name. Record what a\nprofile lacks: the\ngrain one row represents, the index column and its regularity, which columns\ngroup rows, their level counts, what nests in what, each item or covariate's\nunit, missing rate and range, the outcome's distribution and any floor or\nceiling, and every Pearson\u002FSpearman gap that says a relationship is curved.\n\n## Common failure modes\n\n| Symptom | Likely cause | Fix |\n|---|---|---|\n| A \"strong correlation\" that no segment reproduces | pooled Pearson driven by group sizes or a few points | Spearman plus the per-group check above; report the subgroup result |\n| Mean and median disagree wildly, the write-up quotes the mean | heavy tail or an un-flagged outlier | Quote the median with an IQR, flag the tail as its own finding |\n| The reply is 400 lines of `.describe()` and a pairplot | volume substituted for judgement | Cut to findings that change a decision; the rest stays in `run\u002F` |\n| A claim has no `findings.jsonl` row | the number was computed in passing, never recorded | Record it or drop it — `eda-storytelling` rejects unsourced numbers |\n",{"name":5,"description":7},"main","53379a071acd911295e30689207bdb481cbc5466","defaults\u002Fskills\u002Fexploratory-data-analysis",8,"https:\u002F\u002Fgithub.com\u002Fpymc-labs\u002Fdaimon","https:\u002F\u002Fdaimon.pymc-labs.com\u002F","https:\u002F\u002Fgithub.com\u002Fpymc-labs","https:\u002F\u002Fwww.pymc-labs.com","The Bayesian Consultancy","2026-11-14 00:31:23",{"source":8,"skillName":5,"commands":35,"prompt":45,"zipAvailable":46},[36,39,42],{"tool":37,"command":38},"npx","npx skills add pymc-labs\u002Fdaimon --skill exploratory-data-analysis",{"tool":40,"command":41},"bunx","bunx skills add pymc-labs\u002Fdaimon --skill exploratory-data-analysis",{"tool":43,"command":44},"pnpm","pnpm dlx skills add pymc-labs\u002Fdaimon --skill exploratory-data-analysis","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fpymc-labs\u002Fdaimon\u002Fblob\u002Fmain\u002Fdefaults\u002Fskills\u002Fexploratory-data-analysis\u002FSKILL.md",true,[],[49,56,63,70,77,84],{"skillNo":50,"name":51,"nameZh":6,"description":52,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":53,"categories":54,"keywords":55,"publishedAt":19},"01M7929HJTKXW1ZTXENBK958ZE","daimon-context","Use when the user asks about team context, project status, decisions, or discussions that live in Slack or Discord, or names a daimon. Routes the question to the right daimon servers and merges their answers.","https:\u002F\u002Fgithub.com\u002Fpymc-labs\u002Fdaimon\u002Fblob\u002Fmain\u002Fplugin\u002Fskills\u002Fdaimon-context\u002FSKILL.md",[],[],{"skillNo":57,"name":58,"nameZh":6,"description":59,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":60,"categories":61,"keywords":62,"publishedAt":19},"01M7929EP24ZTERFDGV0EH61ZD","qa-echo-skill","QA fixture skill. When asked for the QA echo phrase, reply ECHO-SKILL-OK.","https:\u002F\u002Fgithub.com\u002Fpymc-labs\u002Fdaimon\u002Fblob\u002Fmain\u002Fpackages\u002Ftesting\u002Ffixtures\u002Ftarget53\u002Fcatalog\u002Ffixtures\u002FSKILL.md",[],[],{"skillNo":64,"name":65,"nameZh":6,"description":66,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":67,"categories":68,"keywords":69,"publishedAt":19},"01M7929B143T2Z2K2GWKS9J6AC","workspace-write","Send or draft Gmail and create or edit Google Drive files, Docs and Calendar events through a dedicated shared Workspace identity.","https:\u002F\u002Fgithub.com\u002Fpymc-labs\u002Fdaimon\u002Fblob\u002Fmain\u002Fdefaults\u002Fskills\u002Fworkspace-write\u002FSKILL.md",[],[],{"skillNo":71,"name":72,"nameZh":6,"description":73,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":74,"categories":75,"keywords":76,"publishedAt":19},"01M792966D6TGSTZ3T1G8972W1","workspace-setup","First-time workspace setup and agent-roster operations for a Daimon workspace — working repo, keys, skills, MCP servers, routines, and which changes need an admin.","https:\u002F\u002Fgithub.com\u002Fpymc-labs\u002Fdaimon\u002Fblob\u002Fmain\u002Fdefaults\u002Fskills\u002Fworkspace-setup\u002FSKILL.md",[],[],{"skillNo":78,"name":79,"nameZh":6,"description":80,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":81,"categories":82,"keywords":83,"publishedAt":19},"01M7929307QKVB84Q8F5MV85D4","toggl-edits","Update or bulk edit stopped Toggl Track entries for an explicitly named workspace member using a shared admin key.","https:\u002F\u002Fgithub.com\u002Fpymc-labs\u002Fdaimon\u002Fblob\u002Fmain\u002Fdefaults\u002Fskills\u002Ftoggl-edits\u002FSKILL.md",[],[],{"skillNo":85,"name":86,"nameZh":6,"description":87,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":88,"categories":89,"keywords":90,"publishedAt":19},"01M7928YNB3YG57VQ7GTDZ9274","team-etiquette","Post for other people without wasting their time. Use before posting outside this conversation, tagging people outside it, sharing a document, video, audio or slides with others, or creating or editing a routine that posts for or tags others.","https:\u002F\u002Fgithub.com\u002Fpymc-labs\u002Fdaimon\u002Fblob\u002Fmain\u002Fdefaults\u002Fskills\u002Fteam-etiquette\u002FSKILL.md",[],[]]