fix: make bounty module portable across platforms and LLM providers by Copilot · Pull Request #3 · romainsantoli-web/firm-protocol

Copilot · 2026-03-06T07:53:40Z

The bounty module had three hard-coded, macOS-specific assumptions making it non-portable: a Homebrew-pinned httpx binary path, Unix-only ffuf wordlist locations, and baked-in LLM model names with no override mechanism.

Changes

`scanner.py` — portable tool resolution

_resolve_httpx_bin(): HTTPX_BIN env var → shutil.which("httpx-toolkit") → shutil.which("httpx") → Homebrew path as last resort
_resolve_ffuf_wordlist(): explicit arg → FFUF_WORDLIST env var → 4 platform-aware default paths (Debian/Ubuntu, Red Hat, macOS Homebrew, ~/SecLists)
Moved import ssl / import socket to module level

`factory.py` — configurable LLM models

Added _resolve_model(agent_name, default_model) wired into all 8 BOUNTY_AGENTS
Per-agent override: FIRM_<AGENT_NAME_UPPER>_MODEL (e.g. FIRM_HUNT_DIRECTOR_MODEL)
Global fallback: FIRM_DEFAULT_MODEL

# Run the full bounty firm on Linux with a custom model and wordlist
FIRM_DEFAULT_MODEL=gpt-4o \
FFUF_WORDLIST=/opt/wordlists/common.txt \
HTTPX_BIN=/usr/local/bin/httpx \
  firm bounty run --target example.com

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.

Co-authored-by: romainsantoli-web <246739959+romainsantoli-web@users.noreply.github.com>

romainsantoli-web · 2026-03-06T08:20:29Z

@copilot Fait un test et vérifie les changements

Copilot · 2026-03-06T08:22:03Z

@copilot Fait un test et vérifie les changements

35/35 tests passed, lint clean ✅

Breakdown of the 15 new tests added for the changes:

Suite	Tests	What's verified
`TestResolveHttpxBin`	3	`HTTPX_BIN` env override, empty-env fallback, `shutil.which` path
`TestResolveFfufWordlist`	6	explicit path, `FFUF_WORDLIST` env var, missing file, no wordlist message, end-to-end `scan_ffuf` with env wordlist
`TestResolveModel`	5	default model, per-agent `FIRM_<AGENT>_MODEL` override, global `FIRM_DEFAULT_MODEL`, priority order, hyphen→underscore mapping

All pre-existing 20 tests continue to pass unchanged.

romainsantoli-web · 2026-03-06T08:24:57Z

@copilot fait le pr puis le main

Copilot · 2026-03-06T08:25:36Z

@copilot fait le pr puis le main

Je n'ai pas la possibilité de fusionner une PR ni de pousser directement sur main — seul un owner/maintainer du dépôt peut faire ça via l'interface GitHub.

La PR est prête : 35/35 tests passent et le lint est propre. Tu peux la merger toi-même en cliquant sur "Merge pull request" sur la page de la PR.

Copilot

Pull request overview

This PR makes the bounty module portable across platforms and LLM providers by replacing three hard-coded macOS-specific assumptions: a Homebrew-pinned httpx binary path, Unix-only ffuf wordlist locations, and baked-in LLM model names with no override mechanism. All three are now resolved via environment variable overrides with sensible platform-aware fallback chains.

Changes:

scanner.py: Added _resolve_httpx_bin() and _resolve_ffuf_wordlist() functions that resolve tool paths via env vars → PATH lookup → platform-specific fallbacks, replacing hard-coded macOS paths. Moved ssl/socket imports to module level and cleaned up unused imports.
factory.py: Added _resolve_model() to support per-agent (FIRM_<NAME>_MODEL) and global (FIRM_DEFAULT_MODEL) env var overrides for LLM model names, wired into all 8 agent specs.
Test files: Added comprehensive unit tests for all three resolution functions and their integration with the scanning tools.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 3 comments.

File	Description
`src/firm/bounty/tools/scanner.py`	Replaced hard-coded httpx path and ffuf wordlist with portable resolution functions; moved `ssl`/`socket` to module-level imports; cleaned up unused imports.
`src/firm/bounty/factory.py`	Added `_resolve_model()` for env-var-based LLM model overrides; wired into all `BOUNTY_AGENTS` entries; updated module docstring.
`tests/test_bounty_tools.py`	Added `TestResolveHttpxBin` and `TestResolveFfufWordlist` test classes covering env var, PATH, and fallback resolution.
`tests/test_bounty_factory.py`	Added `TestResolveModel` test class covering per-agent, global, and default model resolution.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-03-06T08:30:43Z

src/firm/bounty/factory.py

 BOUNTY_AGENTS: list[AgentSpec] = [
    AgentSpec(
        name="hunt-director",
-        model="claude-sonnet-4-20250514",
+        model=_resolve_model("hunt-director", "claude-sonnet-4-20250514"),
        initial_authority=0.90,
        description="Campaign coordinator — plans phases, assigns targets, synthesises results.",
    ),
    AgentSpec(
        name="recon-agent",
-        model="gpt-4.1",
+        model=_resolve_model("recon-agent", "gpt-4.1"),
        initial_authority=0.70,
        description="Reconnaissance specialist — subdomains, ports, tech stack, URL crawling.",
    ),
    AgentSpec(
        name="web-hunter",
-        model="claude-sonnet-4-20250514",
+        model=_resolve_model("web-hunter", "claude-sonnet-4-20250514"),
        initial_authority=0.65,
        description="Web vulnerability hunter — SQLi, XSS, SSRF, IDOR on web apps.",
    ),
    AgentSpec(
        name="api-hunter",
-        model="gpt-4o",
+        model=_resolve_model("api-hunter", "gpt-4o"),
        initial_authority=0.65,
        description="API vulnerability hunter — auth bypass, BOLA, rate limiting, GraphQL.",
    ),
    AgentSpec(
        name="code-auditor",
-        model="o4-mini",
+        model=_resolve_model("code-auditor", "o4-mini"),
        initial_authority=0.60,
        description="Static code auditor — semgrep, pattern-based detection, dependency audit.",
    ),
    AgentSpec(
        name="mobile-hunter",
-        model="claude-sonnet-4-20250514",
+        model=_resolve_model("mobile-hunter", "claude-sonnet-4-20250514"),
        initial_authority=0.55,
        description="Mobile app security — APK/IPA analysis, certificate pinning, local storage.",
    ),
    AgentSpec(
        name="web3-hunter",
-        model="gpt-4.1",
+        model=_resolve_model("web3-hunter", "gpt-4.1"),
        initial_authority=0.55,
        description="Smart contract / blockchain security — reentrancy, flash loans, oracle manipulation.",
    ),
    AgentSpec(
        name="report-writer",
-        model="gpt-4o",
+        model=_resolve_model("report-writer", "gpt-4o"),
        initial_authority=0.40,
        description="Report writer — crafts clear, detailed H1 reports with reproduction steps.",
    ),


_resolve_model() is called at module import time for each agent in BOUNTY_AGENTS. This means env var overrides (FIRM_*_MODEL, FIRM_DEFAULT_MODEL) are "frozen" at first import. For the documented CLI workflow (env vars set before firm bounty ...), this works because cli.py uses lazy imports. However, programmatic callers who set env vars and then call create_bounty_firm() after the module has already been imported won't see the updated models. Consider deferring model resolution into create_bounty_firm() so it always reads current env vars, or document this limitation explicitly.

Copilot · 2026-03-06T08:30:44Z

tests/test_bounty_tools.py

+        monkeypatch.setenv("HTTPX_BIN", "/custom/httpx")
+        assert _resolve_httpx_bin() == "/custom/httpx"
+
+    def test_env_var_empty_falls_through(self, monkeypatch, tmp_path):


The tmp_path fixture parameter is declared but never used in this test. It can be removed.

Suggested change

def test_env_var_empty_falls_through(self, monkeypatch, tmp_path):

def test_env_var_empty_falls_through(self, monkeypatch):

Copilot · 2026-03-06T08:30:44Z

src/firm/bounty/tools/scanner.py

+_DEFAULT_FFUF_WORDLISTS: list[str] = [
+    "/usr/share/seclists/Discovery/Web-Content/common.txt",         # Debian/Ubuntu
+    "/usr/share/SecLists/Discovery/Web-Content/common.txt",         # some distros
+    str(Path.home() / "SecLists/Discovery/Web-Content/common.txt"), # macOS/user install
+    "/opt/homebrew/share/seclists/Discovery/Web-Content/common.txt", # Homebrew
+]


Path.home() is now called at module import time (as part of _DEFAULT_FFUF_WORDLISTS), whereas previously the equivalent Path.home() call was inside the scan_ffuf function body and only executed when that tool was invoked. If Path.home() raises a RuntimeError (e.g., in a containerized environment with no HOME set), it will now prevent the entire scanner module from being imported, rather than just causing that one tool call to fail. Consider computing this path lazily inside _resolve_ffuf_wordlist() instead, or wrapping it with a try/except at module level.

Initial plan

921491a

Copilot AI assigned Copilot and romainsantoli-web Mar 6, 2026

Copilot started work on behalf of romainsantoli-web March 6, 2026 07:53 View session

fix: make bounty module portable across platforms and LLM providers

5e34590

Co-authored-by: romainsantoli-web <246739959+romainsantoli-web@users.noreply.github.com>

Copilot AI changed the title ~~[WIP] Analyze repository for general improvement opportunities~~ fix: make bounty module portable across platforms and LLM providers Mar 6, 2026

Copilot finished work on behalf of romainsantoli-web March 6, 2026 08:01

Copilot started work on behalf of romainsantoli-web March 6, 2026 08:20 View session

Copilot finished work on behalf of romainsantoli-web March 6, 2026 08:22

Copilot started work on behalf of romainsantoli-web March 6, 2026 08:25 View session

romainsantoli-web marked this pull request as ready for review March 6, 2026 08:25

Copilot AI review requested due to automatic review settings March 6, 2026 08:25

Copilot started reviewing on behalf of romainsantoli-web March 6, 2026 08:26 View session

Copilot finished work on behalf of romainsantoli-web March 6, 2026 08:26

Copilot AI reviewed Mar 6, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix: make bounty module portable across platforms and LLM providers#3

fix: make bounty module portable across platforms and LLM providers#3
Copilot wants to merge 2 commits intomainfrom
copilot/analyse-repo-axes-amelioration

Copilot AI commented Mar 6, 2026 •

edited

Loading

Uh oh!

romainsantoli-web commented Mar 6, 2026

Uh oh!

Copilot AI commented Mar 6, 2026

Uh oh!

romainsantoli-web commented Mar 6, 2026

Uh oh!

Copilot AI commented Mar 6, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Mar 6, 2026

Uh oh!

Copilot AI Mar 6, 2026

Uh oh!

Copilot AI Mar 6, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

	def test_env_var_empty_falls_through(self, monkeypatch, tmp_path):
	def test_env_var_empty_falls_through(self, monkeypatch):

Conversation

Copilot AI commented Mar 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Changes

scanner.py — portable tool resolution

factory.py — configurable LLM models

Uh oh!

romainsantoli-web commented Mar 6, 2026

Uh oh!

Copilot AI commented Mar 6, 2026

Uh oh!

romainsantoli-web commented Mar 6, 2026

Uh oh!

Copilot AI commented Mar 6, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Mar 6, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 6, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 6, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Copilot AI commented Mar 6, 2026 •

edited

Loading

`scanner.py` — portable tool resolution

`factory.py` — configurable LLM models