feat(skills): malicious skill trust tier enforcement (#1853)#1878
Merged
feat(skills): malicious skill trust tier enforcement (#1853)#1878
Conversation
Extend TrustLevel enforcement in response to empirical study findings (arXiv 2602.06547): 157 confirmed malicious SKILL.md files, 26.1% vulnerability prevalence in community skills. Changes: - Fix QUARANTINE_DENIED tool IDs: replace dead "file_write" with actual FileExecutor IDs (write, edit, delete_path, move_path, copy_path, create_directory) and add memory_save and fetch - Fix execute_tool_call_confirmed delegating to unconfirmed path - Add zeph-tools::patterns module: relocate RAW_INJECTION_PATTERNS and strip_format_chars from zeph-mcp for shared access - Add SkillContentScanner in zeph-skills::scanner: scans skill body at load time using injection patterns, emits WARN with match count; documented as advisory-only (not a security boundary) - Add scan_on_load config flag (default: true) to TrustConfig - Integrate scanner in bootstrap: called for skills below Trusted tier - Add --scan-skills-on-load CLI flag - Add /skill scan TUI command - Update --init wizard with scan_on_load step - Add --migrate-config step for scan_on_load Follow-up: #1876 (MCP tool namespace bypass in QUARANTINE_DENIED)
github-actions
bot
added
documentation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Extends the skill trust enforcement subsystem based on findings from arXiv 2602.06547 (empirical study: 157 confirmed malicious SKILL.md files, 26.1% community skill vulnerability prevalence).
QUARANTINE_DENIEDtool IDs: replace dead"file_write"with actual executor IDs (write,edit,delete_path,move_path,copy_path,create_directory) and addmemory_save,fetchexecute_tool_call_confirmedincorrectly delegating to the unconfirmed pathzeph-tools::patternsmodule: relocateRAW_INJECTION_PATTERNS+strip_format_charsfromzeph-mcpfor shared accessSkillContentScanner(zeph-skills::scanner): scans skill body at load time using injection patterns, emitsWARNwith match count; documented as advisory-onlyscan_on_load = trueconfig flag toTrustConfigTrustedtier--scan-skills-on-loadCLI flag,/skill scanTUI command,--initwizard step,--migrate-configstepTest plan
cargo +nightly fmt --checkpassescargo clippy --workspace --features full -- -D warningspasses (0 warnings)cargo nextest run --config-file .github/nextest.toml --workspace --features full --lib --binspasses (5868 tests)QUARANTINE_DENIEDblocksbash,write,edit,delete_path,move_path,copy_path,create_directory,memory_save,web_scrape,fetchfor Quarantined skillsSkillContentScanneremits WARN on injection pattern matchscan_on_load = falsedisables scannerFollow-up
QUARANTINE_DENIEDdue to namespaced tool IDs (deferred)