feat: OWASP WSTG methodology alignment & TUI live status#328

Open

0xhis wants to merge 41 commits intousestrix:mainfrom

0xhis:prompt-optimization

Commits on Feb 25, 2026

Commits on Mar 9, 2026

Commits on Mar 10, 2026

fix: address Copilot review suggestions
0xhis
committed
fix: revert get_conversation_history copy (memory leak) and remove redundant sleep(0)
0xhis
committed
refactor(prompt): enforce subagent delegation for Phase 1 context gathering
0xhis
committed
fix: address prompt-optimization branch review bugs
0xhis
committed
refactor(prompt): mitigate exploitation phase refusals and simplify agent naming
0xhis
committed
chore: ignore test_run.sh
0xhis
committed
refactor(prompt): update deep scan mode with authorization framing
0xhis
committed
feat(prompt): add mandatory skill assignment triggers for subagent creation
0xhis
committed
fix(agent): mitigate LLM refusals via explicit authorization and attacker perspective constraints
0xhis
committed
fix(agent): add todo list instruction and remove WSTG prefixes from agent names
0xhis
committed
fix(prompt): tighter legal mandate & target infra bypass framing
0xhis
committed
Enhance prompt structure with XML bounding and refusal suppression
0xhis
committed
fix(tool): strictly constrain todo priority values to prevent hallucination
0xhis
committed
fix(agent): fix XML tag nesting and UI rendering issues from PR review
0xhis
committed
fix(agent): stabilize sender attribution and align scan/TUI prompt updates
0xhis
committed
refactor(prompt): condense quick scan mode to baseline-style flow
0xhis
committed
fix(tui): sanitize merged text spans to prevent render crash
0xhis
committed
fix(agent): address review comments for thinking blocks, empty content guard and prompt cleanup
0xhis
committed
fix(tui): sanitize text spans on all single-renderable bypass paths
0xhis
committed
fix(llm): reduce conversation token budget to 80k to prevent exceeding model context limit
0xhis
committed
fix(llm): include system prompt tokens in memory compressor budget
0xhis
committed
fix(llm): handle malformed function/parameter open tags from GLM-5
0xhis
committed