Repository navigation
quiet positioning: README pivot + fixed-context dedup (budget re-frozen at 4580) - #17
Merged
Merged
Conversation
Measures the per-session fixed context (SessionStart hook payload + skill metadata) of coding-discipline and superpowers with symmetric rules, so the README comparison table can be reproduced by anyone. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…d-cost table Measured against superpowers v6.1.1 (post its v6.1.0 bootstrap trim), fixed injection is a tie: 5,230 vs 5,556 chars (~1.3k tokens each). Drop the implied "lighter" angle; state that the difference is how often the plugin intervenes and that interventions add no subagents, plan files, or artifacts. Adds the measured table with a reproduction link and a commented placeholder for two demo clips. Mirrored in README.zh-CN.md. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Remove four pure duplications from the fixed injection (Tier A only, no trigger-semantics change): - primer: drop the "Do not merely read SKILL.md" paragraph (the hook already appends a host-specific invocation note on both supported hosts) and the human-facing H1 parenthetical - git-flow description: drop the project-guide deference clause (covered by the primer's global precedence rule and the skill body) - systematic-debugging description: drop the emergency-containment sentence (covered verbatim by the skill body's hard rule) Give the hook's unknown-host branch a generic invocation note so exotic hosts keep the guidance the primer sentence used to carry. Re-freeze FIXED_CONTEXT_BUDGET at the new 4580-character baseline (was 4944) and update the CLAUDE.md hard boundary to match. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Post-trim measurement: hook payload 1,474 chars, skill metadata 3,392 chars, total 4,866 chars (~1.2k tokens) vs superpowers v6.1.1 at 5,556. Label the column (main) instead of v0.8.0 since the trimmed numbers are not the v0.8.0 release's, and reword the frozen-baseline note in both READMEs to 4580. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
chipfighter
changed the title
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Two-part scope, one campaign — say the positioning and be it:
1. README pivot (both languages): intervention frequency, not "lighter".
Measured fixed injection against superpowers v6.1.1 with symmetric rules
(everything the SessionStart hook injects + skill metadata the host loads;
bodies excluded on both sides). Result is a near-tie, so the README now
leads with the honest claim: fixed cost is a wash; the differentiator is
how often a plugin inserts itself and what it spawns when it does.
Reproducible via experiments/context-measure/measure_fixed_context.py.
2. Fixed-context dedup (Tier A only — pure duplication, no
trigger-semantics change):
appends a host-specific invocation note) and the human-facing H1
parenthetical
by the primer's global precedence rule + the skill body)
sentence (covered verbatim by the body's hard rule)
Behavior-summary sentences in descriptions (Tier B) are deliberately NOT
trimmed: the routing surface stays stable while real users generate
trigger feedback.
Measured (2026-07-17, post-trim)
Evidence
tests/test-plugin-metadata.pypasses (budget 4580 is an exact fit)tests/test-hooks.shpasses (bash suite)tests/test-windows-hook.ps1passespresent, removed sentence gone
Left before merge
🤖 Generated with Claude Code