📦

Audit History

executor-planning - 1 audit

Oct 5, 2026, 06:00 PM

Most of the 221 static findings are Markdown formatting, legitimate companion references, or routine planning commands. Two temporary-file findings remain confirmed, and a semantic finding identifies overly broad Git staging. No evidence found of prompt injection or exfiltration; companion script implementations are outside the reviewed package.

Files scanned

773

Lines analyzed

Review items

False positives ignored

Confirmed security concerns (1)

Medium

SKILL.md:710-715

Broad Git Staging Can Include Unrelated Changes

The handoff runs 'git add docs/executor/INIT-0004-cloud-tenant-cells' before committing. Directory-wide staging can include unrelated or sensitive initiative files beyond the intended plan, index, and specification updates.

The example explicitly stages the entire initiative directory instead of the three intended document updates. Actual exposure depends on other changes in that directory.

Capability review items (2)

These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.

Medium

SKILL.md:581

Temp directory access

"$n" /tmp/brief-check.md >/dev/null || echo "TASK $n FAILS"

The example writes each brief to a predictable shared temporary pathname without showing exclusive creation. Concurrent runs or precreated links could corrupt or redirect output.

Medium

SKILL.md:582

Temp directory access

grep -m1 '^\*\*Task:\*\*' /tmp/brief-check.md

The loop reads the same shared temporary file even after extraction failure. Stale output or another concurrent run could substitute the task identity being verified.

Risk Factors

⚙️ External commands (50)

📁 Filesystem access (21)

Audited by: codex

← Back to Skill