Prompt Compression
PendingStatic analysis audit pending.
Overview
No static analysis result has been recorded yet. Pattern checks will appear here once the artifact has been analyzed.
Findings (0)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
Compressed wording could unintentionally change how agents or skills behave across a repository.
The default target list includes broad agent, skill, and system-prompt files, so one run can change multiple files that affect future agent behavior.
"targets": [".ai-workflow/skills/**/SKILL.md", "skills/**/SKILL.md", "scripts/ralph/CLAUDE.md", "claude.md", "openai.md", ".cursorrules", "system_prompt.md"]
Review and narrow the target list before running, inspect diffs afterward, and use the undo path if behavior changes unexpectedly.
Important instructions may become less clear or slightly change meaning after automated compression.
Balanced mode applies rule-based instruction compression, including deleting some instruction-introducing phrases; this is central to the purpose but can affect clarity if applied to safety- or approval-related prompts.
"balanced": ["remove_filler", "fix_whitespace", "deduplicate", "compress_instructions", "shorten_examples"], ... "You are required to": ""
Avoid applying this blindly to security-critical prompts; review diffs before committing or using compressed prompts in production loops.
The documented install commands may fail, and any separately obtained installer script would not be covered by this review.
The README instructs users to run installer scripts, but the supplied manifest contains no scripts directory or installer files, so those runnable files were not available for review.
powershell -NoProfile -ExecutionPolicy Bypass -File skills/trinity-compress/scripts/install.ps1 -RepoPath .
Use only reviewed files from the package, verify any installer script before running it, or wait for a package that includes the referenced scripts.
Old prompt content could remain in the repository as backup files and might be accidentally shared if not ignored.
Backup files intentionally persist copies of prompt/instruction files, which may include private project context or sensitive agent instructions.
Creates **.bak** backups and supports instant undo.
Keep *.bak files ignored, avoid placing secrets in prompt files, and clean backups when they are no longer needed.