UNPKG

@crossplatformai/skills

Version:

Reusable Agent Skills for CrossPlatform.ai projects.

82 lines (61 loc) • 4.23 kB
--- name: do-not-abbreviate description: Require complete words in newly authored human-facing copy, names, labels, documentation, messages, and explanatory comments, with authorization before a shortened form; do not rewrite existing exact text, machine-facing tokens, or user-provided strings. --- # Do Not Abbreviate ## Use This Skill When Use this skill when creating or revising human-facing content, including visible copy, visible names and labels, documentation, explanatory messages, and explanatory comments. Apply the rule to newly authored sentences, additions, names, and labels. Existing text is preserved unless the authorized user explicitly asks for a change. This skill owns shortened-form policy. Load `documentation-writing` separately when the task also needs guidance about evergreen or change-oriented document framing. ## Default Write complete words and phrases in newly authored prose, names, and labels. Avoid abbreviations, acronyms, initialisms, clipped terms, and internal shorthand. Do not treat commonness, team convention, previous repository usage, or presumed reader familiarity as permission. A shortened form that is familiar to the team or common in the technology industry still needs authorization. ## Authorization Gate When a shortened form would be useful: 1. Stop before writing it. 2. Ask the authorized user for explicit permission. 3. Identify all four parts of the request: - **Full term:** the complete word or phrase. - **Shortened form:** the exact proposed abbreviation, acronym, initialism, clipped term, or internal shorthand. - **Reason:** exactly one permitted reason. - **Scope:** every name, label, message, comment, or document location where the permission applies. 4. Use the shortened form only after approval and only within the approved scope. Otherwise, write the full term. Only the authorized user can grant permission. Research, another agent, repository convention, and prior approval for a different scope cannot substitute for that instruction. Keep an approved shortened form consistent within its approved scope; do not carry it into a different scope without another authorization. Only these two reasons may authorize a shortened form: 1. **Explicit character-space limit:** the destination has an identified maximum number of characters. A real limit does not grant automatic permission; prior authorization is still required. 2. **Deliberate obfuscation:** the shortened form intentionally hides or distinguishes information from the audience, such as a deliberately opaque visible release name. Conciseness, speed, clarity, common usage, team convention, previous repository usage, presumed familiarity, or research is not a third reason. Research may identify the full term, but it cannot authorize the shortened form. If the authorized user does not approve one of the two permitted reasons, fail fast by writing the full term or asking for the missing authorization. Newly created visible names and labels remain in scope. They may use a shortened form for deliberate obfuscation only when the authorized user explicitly approves that purpose and the scope. ## Preserve Existing Exact Text Do not expand or reword existing exact tokens or source-of-truth text merely to follow this rule. Preserve existing identifiers, commands, protocol tokens, web addresses, file formats, quotations, and user-provided strings. When adding to existing content, leave the existing material unchanged and apply this skill to the new material. If existing shorthand conflicts with new full-word prose, surface the mismatch without silently rewriting the existing source. ## Boundaries Ordinary grammatical contractions and symbols required by a language, format, protocol, or user interface are outside this rule. Code identifiers that are not shown to people, commands, protocol tokens, web addresses, file formats, quotations, and user-provided strings are also outside the newly authored human-facing scope. This is writing guidance, not a linter, detector, runtime enforcement mechanism, glossary, or universal language guide. Do not rewrite unrelated existing content or infer permission from an earlier use.