Arc Shield

SuspiciousAudited by ClawScan on May 10, 2026.

Overview

Arc Shield is a coherent local sanitizer, but its strict mode can still print the unredacted message before failing, which can leak the secrets it is meant to block.

Review or patch the strict-mode behavior before installing this as a pre-send safety control. Until then, treat it as a local audit/redaction helper only, avoid piping its strict-mode stdout to external channels, and verify all sample secrets are fake.

Findings (3)

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.

What this means

If this is placed in a pipeline, terminal log, or send wrapper, secrets may still be exposed even though the command exits with a blocking status.

Why it was flagged

Strict mode prints the full input before deciding whether critical secrets were found, so a secret-containing message can be emitted to stdout before the command fails.

Skill content
elif [[ "$MODE" == "strict" ]]; then echo "$INPUT"; if [[ $FOUND_CRITICAL -gt 0 ]]; then
Recommendation

Do not pipe --strict output directly to any external send/log path until fixed; change strict mode to suppress output on failure or only emit redacted text, and add tests for secret-containing strict-mode stdout.

What this means

Users may trust the documented command as a safe pre-send filter and accidentally expose unredacted secrets.

Why it was flagged

The documentation presents --strict as a blocking safeguard for external messages, but the implementation shown above can output the original message before blocking.

Skill content
# Block if critical secrets found (use before external messaging) echo "Message text" | arc-shield.sh --strict || echo "BLOCKED"
Recommendation

Update the documentation to clearly state stdout/stderr behavior, show safe wrappers that discard stdout on block, and avoid claiming blocking protection until strict mode is corrected.

What this means

Users have less provenance information for verifying they installed the intended package and supporting files.

Why it was flagged

The reviewed files are local scripts, but the registry metadata does not provide a clear source or install specification, while the docs describe git/chmod setup.

Skill content
Source: unknown; Homepage: none; No install spec — this is an instruction-only skill.
Recommendation

Install only from a verified repository or signed bundle, and confirm required files such as scripts and documented configuration are present before use.