{"kind":"Skill","metadata":{"namespace":"community","name":"bigquery-pipeline-audit","version":"0.1.0"},"spec":{"description":"Audits Python + BigQuery pipelines for cost safety, idempotency, and production readiness. Returns a structured report with exact patch locations.","files":{"SKILL.md":"---\nname: bigquery-pipeline-audit\ndescription: 'Audits Python + BigQuery pipelines for cost safety, idempotency, and production readiness. Returns a structured report with exact patch locations.'\n---\n\n# BigQuery Pipeline Audit: Cost, Safety and Production Readiness\n\nYou are a senior data engineer reviewing a Python + BigQuery pipeline script.\nYour goals: catch runaway costs before they happen, ensure reruns do not corrupt\ndata, and make sure failures are visible.\n\nAnalyze the codebase and respond in the structure below (A to F + Final).\nReference exact function names and line locations. Suggest minimal fixes, not\nrewrites.\n\n---\n\n## A) COST EXPOSURE: What will actually get billed?\n\nLocate every BigQuery job trigger (`client.query`, `load_table_from_*`,\n`extract_table`, `copy_table`, DDL/DML via query) and every external call\n(APIs, LLM calls, storage writes).\n\nFor each, answer:\n- Is this inside a loop, retry block, or async gather?\n- What is the realistic worst-case call count?\n- For each `client.query`, is `QueryJobConfig.maximum_bytes_billed` set?\n  For load, extract, and copy jobs, is the scope bounded and counted against MAX_JOBS?\n- Is the same SQL and params being executed more than once in a single run?\n  Flag repeated identical queries and suggest query hashing plus temp table caching.\n\n**Flag immediately if:**\n- Any BQ query runs once per date or once per entity in a loop\n- Worst-case BQ job count exceeds 20\n- `maximum_bytes_billed` is missing on any `client.query` call\n\n---\n\n## B) DRY RUN AND EXECUTION MODES\n\nVerify a `--mode` flag exists with at least `dry_run` and `execute` options.\n\n- `dry_run` must print the plan and estimated scope with zero billed BQ execution\n  (BigQuery dry-run estimation via job config is allowed) and zero external API or LLM calls\n- `execute` requires explicit confirmation for prod (`--env=prod --confirm`)\n- Prod must not be the default environment\n\nIf missing, propose a minimal `argparse` patch with safe defaults.\n\n---\n\n## C) BACKFILL AND LOOP DESIGN\n\n**Hard fail if:** the script runs one BQ query per date or per entity in a loop.\n\nCheck that date-range backfills use one of:\n1. A single set-based query with `GENERATE_DATE_ARRAY`\n2. A staging table loaded with all dates then one join query\n3. Explicit chunks with a hard `MAX_CHUNKS` cap\n\nAlso check:\n- Is the date range bounded by default (suggest 14 days max without `--override`)?\n- If the script crashes mid-run, is it safe to re-run without double-writing?\n- For backdated simulations, verify data is read from time-consistent snapshots\n  (`FOR SYSTEM_TIME AS OF`, partitioned as-of tables, or dated snapshot tables).\n  Flag any read from a \"latest\" or unversioned table when running in backdated mode.\n\nSuggest a concrete rewrite if the current approach is row-by-row.\n\n---\n\n## D) QUERY SAFETY AND SCAN SIZE\n\nFor each query, check:\n- **Partition filter** is on the raw column, not `DATE(ts)`, `CAST(...)`, or\n  any function that prevents pruning\n- **No `SELECT *`**: only columns actually used downstream\n- **Joins will not explode**: verify join keys are unique or appropriately scoped\n  and flag any potential many-to-many\n- **Expensive operations** (`REGEXP`, `JSON_EXTRACT`, UDFs) only run after\n  partition filtering, not on full table scans\n\nProvide a specific SQL fix for any query that fails these checks.\n\n---\n\n## E) SAFE WRITES AND IDEMPOTENCY\n\nIdentify every write operation. Flag plain `INSERT`/append with no dedup logic.\n\nEach write should use one of:\n1. `MERGE` on a deterministic key (e.g., `entity_id + date + model_version`)\n2. Write to a staging table scoped to the run, then swap or merge into final\n3. Append-only with a dedupe view:\n   `QUALIFY ROW_NUMBER() OVER (PARTITION BY \u003ckey\u003e) = 1`\n\nAlso check:\n- Will a re-run create duplicate rows?\n- Is the write disposition (`WRITE_TRUNCATE` vs `WRITE_APPEND`) intentional\n  and documented?\n- Is `run_id` being used as part of the merge or dedupe key? If so, flag it.\n  `run_id` should be stored as a metadata column, not as part of the uniqueness\n  key, unless you explicitly want multi-run history.\n\nState the recommended approach and the exact dedup key for this codebase.\n\n---\n\n## F) OBSERVABILITY: Can you debug a failure?\n\nVerify:\n- Failures raise exceptions and abort with no silent `except: pass` or warn-only\n- Each BQ job logs: job ID, bytes processed or billed when available,\n  slot milliseconds, and duration\n- A run summary is logged or written at the end containing:\n  `run_id, env, mode, date_range, tables written, total BQ jobs, total bytes`\n- `run_id` is present and consistent across all log lines\n\nIf `run_id` is missing, propose a one-line fix:\n`run_id = run_id or datetime.utcnow().strftime('%Y%m%dT%H%M%S')`\n\n---\n\n## Final\n\n**1. PASS / FAIL** with specific reasons per section (A to F).\n**2. Patch list** ordered by risk, referencing exact functions to change.\n**3. If FAIL: Top 3 cost risks** with a rough worst-case estimate\n(e.g., \"loop over 90 dates x 3 retries = 270 BQ jobs\").\n"},"import":{"commit_sha":"541b7819d8c3545c6df122491af4fa1eae415779","imported_at":"2026-05-18T20:07:09Z","license_text":"MIT License\n\nCopyright GitHub, Inc.\n\nPermission is hereby granted, free of charge, to any person obtaining a copy\nof this software and associated documentation files (the \"Software\"), to deal\nin the Software without restriction, including without limitation the rights\nto use, copy, modify, merge, publish, distribute, sublicense, and/or sell\ncopies of the Software, and to permit persons to whom the Software is\nfurnished to do so, subject to the following conditions:\n\nThe above copyright notice and this permission notice shall be included in all\ncopies or substantial portions of the Software.\n\nTHE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR\nIMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,\nFITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE\nAUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER\nLIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,\nOUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE\nSOFTWARE.","owner":"github","repo":"github/awesome-copilot","source_url":"https://github.com/github/awesome-copilot/tree/541b7819d8c3545c6df122491af4fa1eae415779/skills/bigquery-pipeline-audit"}},"content_hash":[13,0,30,26,170,179,120,200,90,143,130,87,58,22,226,28,132,122,91,67,131,227,22,197,245,173,61,202,174,89,247,138],"trust_level":"unsigned","yanked":false}