Memory Compress
PassAudited by ClawScan on May 1, 2026.
Overview
This appears to be a transparent local markdown memory-log compressor, but users should choose file paths carefully and review summaries before adding them to long-term memory.
This skill looks coherent and local-only. Before installing, be comfortable with it reading memory log files and writing summaries, run it only on intended paths, review the compressed output before appending to MEMORY.md, and do not rely on the summary as truly lossless.
Findings (4)
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.
If invoked with the wrong path, the skill could summarize an unintended local markdown file or write the summary somewhere unexpected.
The script reads and writes caller-supplied paths and creates output directories. This is expected for a local log compressor, but paths are not confined to the documented workspace/memory area.
const logPath = path.resolve(WORKSPACE, DAILY_LOG); ... fs.mkdirSync(outputDir, { recursive: true }); ... fs.writeFileSync(path.resolve(OUTPUT), compressed, 'utf-8');Use explicit relative paths under the intended memory directory, avoid absolute or parent-directory paths, and consider adding path validation before broader use.
A user may not realize before installation that running the skill requires a working Node.js runtime.
The skill documentation runs a Node script, so the runtime prerequisite is under-declared in metadata. The included source uses only built-in modules, so this is an install clarity issue rather than hidden package behavior.
Required binaries (all must exist): none; No install spec — this is an instruction-only skill.
Declare Node.js as a required binary or clearly document the runtime requirement in registry metadata.
Bad or misleading summaries could influence future agent behavior after being appended to MEMORY.md.
The skill is designed to move summarized log content into persistent agent memory. The review step helps, but inaccurate or instruction-like log content could be preserved if not checked.
2. Review for accuracy; 3. Append: cat /tmp/compressed.md >> MEMORY.md
Always review the compressed output before appending it to persistent memory, and keep the raw logs available for correction.
A user may over-trust the compressed memory and miss details that were not captured.
The skill makes very strong lossless-compression claims, while the documented approach is heuristic extraction of key sections and top items. Users should treat the output as a summary, not a guaranteed complete record.
4-8x compression ratio, zero information loss ... No data loss, ever.
Reword the claim to describe best-effort summarization, and verify important events, decisions, and todos against the raw log.