Exploratory agent dossier · verified 22 Sep 2026
Loopmark Agent: a constrained local drafting pilot.
Loopmark is the narrow-agent-app option in this compilation. The checked evidence supports a Python/LangGraph CLI with local JSON persistence and draft-oriented marketing tools; it does not support a production autonomy claim.
Scope and dependencies
Boundary. The README describes content, email, funnel and audience-research tools, optional scheduler/posting, and local persistence.
License. The checked LICENSE is MIT, an OSI-style license classification. This says nothing on operational maturity.
The documented setup lists Python 3.10+, an OpenAI key and optional LangSmith plus X, LinkedIn and Buffer credentials. Models, social APIs and tracing are separate cost and data-handling decisions.
Why it remains exploratory
The checked open issues list managed retries/alerts, RBAC/approvals, Google Ads, LinkedIn Campaign Manager/follower analytics and Meta Ads as open or planned. Do not translate those roadmap items into shipped capabilities. No dedicated security-policy or advisory process was found in the reviewed evidence.
Conditional decision
Safe first test: use a disposable local data directory and low-spend test key, leave social credentials unset, create draft content, inspect the local JSON and delete the folder.
Maintenance signal and limit
The checked main history showed a latest visible commit on 31 August 2026; the README says tests/CI run and warns against committing secrets. These are thin signals relative to the operating gaps, so the compilation treats Loopmark as an exemplar for cautious exploration rather than a launch recommendation.