Abbreviations

abbreviations — naming & terse-writing dictionary

Single source of truth for standard abbreviations in names (vars / fns / classes / CSS classes / tokens) and terse text (comments, commit msgs, logs).

This policy, the canonical dictionary.tsv, and your local capture buffer ${ABBR_LOCAL:-$HOME/.config/abbr/local.tsv} (if it exists) are auto-injected into every session — incl. subagents — by this plugin's SessionStart hook (hooks/abbr-inject.sh). All of it is already in your context; you don't open the files to look an abbr up. Never duplicate the dict anywhere else.

Policy

• Only mainstream, at-a-glance abbrs. Ones any dev recognizes on sight. No private / project / company jargon in the canonical dict (shared across all projects, no personal data). Dubious / local / stack-specific → spell the word out in full, don't add it.
• Names self-explain. A var / fn / class / CSS class / token must be clear from its name alone; if you must comment what a name means, rename it.
• Comments explain "why", not "what". Minimal comments — only a nontrivial edge case, a bug workaround, a hidden invariant. A good name needs no "what".

Disambiguation

The 2-col data file can't resolve ambiguity — pick the explicit form, don't guess:

• res — if a module already uses res=resource, use resp for response.
• acc — accumulator vs account: on any risk of confusion, spell it out.
• tx — transaction (the default); in net/radio code spell out transmit/text.
• mux — mutex vs multiplexer (HTTP router): qualify by context.
• ns — namespace vs nanosecond: pick the explicit name.
• ref vs ptr — ref = logical reference/key, ptr = pointer.

Growing the dictionary

Coined or hit a genuinely mainstream abbr missing from both files → append it to the capture buffer, and periodically promote buffer rows into canonical. Full flow (buffer format, append snippet, promotion steps) → growing.md.