GHSA-74p7-6h78-gw8p: skillctl: argument injection, path traversal in --dest, FIFO/device DoS, hardlink exfiltration, and commit-trailer forgery
Following the path-safety patches in GHSA-wx3m-whqv-xv47 (v0.1.2), a comprehensive multi-angle audit surfaced five further vulnerabilities, now patched in v0.1.3:
source_shaargument injection ingit ls-tree(CRITICAL).InstalledSkill.source_shadeserialized from.skills.toml(committed, PR-mergeable) flowed unvalidated intogit ls-tree -r -z <refspec> -- <path>. Because the refspec sits before--, an attacker who slipped a malicious.skills.tomlinto a PR could setsource_sha = "--name-only"/--abbrev=0/--output=…and corrupt the diff classifier (which drivespull/pushdestructive decisions), or forge a divergence state to trickpush --on-divergence overwriteinto clobbering the wrong content.- FIFO / device / socket denial-of-service in
copy_dir_all(CRITICAL). The file-type branch only checkedis_dir()/is_symlink(); a FIFO inside a skill folder fell through tofs::copy, which blocks indefinitely waiting for a writer. A character device like/dev/zerowould read until OOM. Reachable onskillctl addagainst any adversarial library. add --destarbitrary-directory wipe in agent mode (HIGH).--destaccepted absolute paths and..traversal without validation, so a single invocationskillctl add --dest /Users/victim/.ssh --on-conflict overwrite --skill <maliciously-named>wouldremove_dir_allarbitrary directories — no.skills.tomlround-trip required. Reachable in any non-interactive / agent-driven workflow where flag values may be attacker-supplied.- Commit-message trailer forgery via skill names (HIGH). Skill names were spliced verbatim into
git commit -m "update skill: <name>"and into thecommit.messagefield of--jsonoutput. A skill namedfoo\nCo-Authored-By: evil@xproduced a forged commit trailer that downstream tooling (Linear, GitHub commit-bot, release-notes scrapers) treats as real authorship metadata. - Hardlink exfiltration via the round-trip (HIGH).
fs::symlink_metadatareports a regular file for hardlinks (shared inode), andfs::copyreads the target content. An untrusted agent writing<project>/my-skill/dataas a hardlink to~/.ssh/id_rsawould have shipped the SSH key content to the (possibly public) library on the nextskillctl pushordetect.
References
Code Behaviors & Features
Detect and mitigate GHSA-74p7-6h78-gw8p with GitLab Dependency Scanning
Secure your software supply chain by verifying that all open source dependencies used in your projects contain no disclosed vulnerabilities. Learn more about Dependency Scanning →