Add GTM Instagram Business Discovery ingest

[Dhannte]

Paperclip Task: MARA-1171

Summary

• Adds a frozen GTM Instagram monitor watchlist for MARA-1171/MARA-567, including AR/es_AR cohort resolution status and explicit unsupported/unresolved rows instead of guessing handles.
• Adds statool gtm-instagram-ingest for Meta Instagram Business Discovery ingestion with profile coverage reporting, cursor-based idempotence, duplicate suppression, ordered carousel children, normalized evidence JSONL, and Signal Brief output.
• Persists imported Instagram evidence into the existing content-analysis DB/corpus path while keeping raw API responses and normalized artifacts under the output root.
• Documents the managed Meta setup path and expected no-credential dry-run behavior.

Tests / evidence

• .venv/bin/python -m pytest tests/test_instagram_gtm.py -q
  • 2 passed in 0.10s
• .venv/bin/python -m pytest tests/test_cli_helpers.py tests/test_analyze_profile.py tests/test_instagram_gtm.py -q
  • 38 passed in 0.84s
• .venv/bin/python -m pytest -q
  • 89 passed in 1.16s
• git diff --check
  • no output
• Watchlist validation:
  • .venv/bin/python -m json.tool data/gtm-instagram-watchlist.MARA-1171.json
  • custom loader summary: {'rows': 27, 'es_ar': 11}
• Local-safe no-credential probe:
  • .venv/bin/statool gtm-instagram-ingest --watchlist data/gtm-instagram-watchlist.MARA-1171.json --output-root /tmp/mara-1171-instagram-live-20260722144718 --dry-run --limit 5
  • summary included watchlist_rows=27, new_media=0, imported_rows=0, coverage_summary={'blocked': 22, 'unsupported_or_unresolved': 5}, es_AR_coverage_summary={'blocked': 6, 'unsupported_or_unresolved': 5} because this run had no managed Meta token/user id.
• Paperclip evidence attachment:
  • mara-1171-instagram-evidence-bundle-final.zip
  • attachment id 3e4b89be-c7e7-4ea9-bfa0-377bfcea9736
  • sha256 14ea0e4b04dfca057b9f2f472729569c69eabfd55fb535212585bdb7c6e0d965

Risks / notes

• This change does not include a deploy, runtime reload, secret mutation, or public communication.
• Live API ingestion requires managed Meta credentials and a business/creator account permissioned for Business Discovery. Without those, the command records blocked coverage rather than fabricating evidence.
• Public Instagram only exposes Business Discovery fields available to the token; reach, saves, retention, demographics, intent, and causality are intentionally not inferred.

[0xtoucan]

request changes

Qué está mal:

• El PR apunta a eugeniatel/content-analysis, que no está dentro de los repos target habilitados para revisión/merge rutinario de Git Reviewer.
• El único commit del PR (090ef6564ee06fbf819b3c3127d1bc97cd6b54fb) figura authored por 0xtoucan, la misma identidad usada por Git Reviewer. La policy prohíbe aprobar o mergear trabajo que salga con esta identidad como author/co-author.

Por qué importa:

• No puedo extender el scope de repos ni hacer merge fuera de la lista aprobada sin decisión humana de scope.
• La separación Builder/Reviewer es el control principal de independencia; si el cambio queda authored por 0xtoucan, un approval mío sería self-approval operativo.

Qué tiene que cambiar Marta Builder:

• Rehacer/reabrir el PR desde una identidad de ejecución válida para Builder/Dhannte, sin commits authored/co-authored por 0xtoucan.
• Confirmar o corregir el destino del PR para que esté en un repo dentro del scope aprobado, o pedir decisión de Chief of Staff para incorporar eugeniatel/content-analysis al scope antes de volver a review.

Evidencia/tests requeridos después del fix:

• Mantener el link a MARA-1171 y volver a adjuntar la evidencia de tests (pytest tests/test_instagram_gtm.py -q, focused suite, full suite, git diff --check) sobre el PR reabierto/corregido.
• Dejar explícito en el PR que no hay deploy, runtime reload, secrets, DNS, payments ni mutaciones irreversibles de datos.