Detects subagent capability escapes by comparing the tool being invoked against the skill's declared allowed-tools list. Use when the developer asks about subagent escape, sees a tool fired from an unexpected skill, or audits delegation.md compliance. Auto-fires on PreToolUse. Best-effort observability only — does not block.
How this skill is triggered — by the user, by Claude, or both
Slash command
/hydra-capability-fence:fence-awarenesshaikuThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
<purpose>
<output_format>
Tool fired: <tool_name> Declared whitelist: Classification: <genuine-escape | honest-omission | injection-symptom>
Recommended next step:
This is observability only; the tool call was NOT blocked. Real enforcement requires a harness-level capability overlay (F-050). </output_format>
1. NEVER claim the hook blocked anything — it cannot. 2. NEVER edit SKILL.md frontmatter without showing the diff first. 3. ALWAYS cross-reference `wixie/../vis/packages/core/conduct/delegation.md` for the canonical per-role tool whitelist. 4. ALWAYS note that absence of evidence in fence-log.ndjson is not evidence of compliance — the hook only fires when a SKILL.md is in scope.<failure_modes>
Guides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.
npx claudepluginhub enchanter-ai/hydra --plugin capability-fence