fix(agent-scan): register 4 new detection skills in _DETECTION_SKILLS

[boy-hack]

Summary

Four detection skills added in PR #444 were never registered in _DETECTION_SKILLS in agent-scan/core/agent.py, so they were silently never executed during agent scans.

Root Cause

PR #444 added 4 new skill files under agent-scan/prompt/skills/ but did not update the _DETECTION_SKILLS list in agent.py. This list controls which skills are actually spawned as workers during Stage 2 of an agent scan.

Fix

Added the 4 missing skills to _DETECTION_SKILLS:

_DETECTION_SKILLS: List[str] = [
    "data-leakage-detection",
    "tool-abuse-detection",
    "indirect-injection-detection",
    "authorization-bypass-detection",
    # Newly registered (were present but not executed):
    "direct-injection-detection",
    "file-path-traversal-detection",
    "hardcoded-secret-detection",
    "memory-poisoning-detection",
]

Impact

After this fix, agent scans will detect:

• Direct Prompt Injection (direct-injection-detection) — ASI01
• File Path Traversal (file-path-traversal-detection) — ASI05
• Hardcoded Secrets (hardcoded-secret-detection) — ASI06
• Memory Poisoning (memory-poisoning-detection) — ASI06

These are critical security detection capabilities that were previously available but silently skipped.

Reference: PR #429 (by @DevamShah) correctly demonstrated the pattern — both SKILL.md + agent.py registration in the same PR.