Editorial Policy¶

Voice, citation, safety, and review standards. Enforced by the Documentation Agent at finalize time and by the Publishing Agent at preflight.

Voice¶

• Plain, direct, declarative. Avoid hype, marketing prose, hedge words.
• Second person for instructional content; third person for narrative.
• Active voice unless passive serves clarity.
• Define jargon at first use; link rather than redefine if the term has a wiki page.

Clarity standards¶

• Lead with the takeaway. Then the why. Then the how.
• One idea per paragraph; one paragraph per code block where reasonable.
• Section headings reflect the reader's question, not the writer's section number.
• Show, don't tell: examples and verifiable claims preferred over assertions.

Citation rules¶

• Every non-obvious factual claim has a source reference (source_references[] in the content schema).
• Direct quotes use a verbatim quote and a source.
• Internal references link to the canonical wiki page; do not duplicate facts across pages.
• External sources are stable (commit-pinned, archive.org-saved if at risk).

Safety constraints¶

• No claims that imply security/legal/medical/financial advice unless explicitly authored as such with appropriate disclaimers.
• No personal information about individuals without consent (SPB-01).
• No leaked secrets, tokens, or credentials in any content (SPB-09).
• No code that downloads/executes anything outside the project's published artifacts.
• Adversarial-LLM outputs from the security sandbox never ship in public content unless redacted and reviewed.

Prohibited claims¶

• Claims of guaranteed outcomes or compliance ("100% secure", "always works").
• Disparagement of identifiable people or organizations.
• Predictions presented as facts.

Review requirements by content type¶

Failure handling¶

• Editorial check failure blocks finalization.
• Re-review after every substantive content change.
• Non-substantive (typo, formatting) changes do not require full re-review but must be logged.