[Gap] RedditTool + standardized Claw/demo path (ALLOW_WRITE, claw scripts parity)

[Dhivya-Bharathy]

Executive summary

A RedditTool exists in praisonai_tools, enabling basic Reddit automation flows. Hermes demos and OpenClaw bridges often emphasize deterministic claw scripts (run_claw.ps1, MCP manifests, health checks) pulling a frozen tool bouquet.

Today the story for Reddit + Claw is inconsistent: unclear which entrypoint, venv, HERMES_ONLY_TOOLS subset, or OAuth vs read-only endpoints wins in production-ish demos.

This issue closes the glue layer gap between RedditTool implementation and standardized claw/demo packaging.

Repos: praisonai-tools (implementation surface) + demo/monorepo scripts (run_claw.ps1 family) per ISSUE_INDEX.md item #10.

---

Architecture map

┌─────────────────┐       ┌──────────────────┐
│  RedditTool     │       │ Claw MCP bridge  │
│  (praison_tools)│◄─────►│ / HTTP surface   │
└────────┬────────┘       └────────┬─────────┘
         │ oauth / httpx           │ process mgmt + env injection
         ▼                         ▼
 ┌────────────────┐       ┌──────────────────┐
 │ Reddit OAuth   │       │ Developer Win/UX │
 │ reddit_* env   │       │ scripts + docs   │
 └────────────────┘       └──────────────────┘

Gap: middle column manifest + script parity incomplete.

---

Workstreams

W1 — Tool surface audit

Catalog:

• Implemented operations (subscribe? post? read subreddit?)
• Missing Hermes-aligned ops (vote? save? moderator actions?) → defer unless security-reviewed

Expose explicit JSON schemas reflecting scopes Reddit grants.

W2 — Claw standard path

Produce canonical snippet:

# illustrative
tools_allowlist:
  - reddit_get_hot
  - reddit_post_comment # if permitted
environment:
  REDDIT_CLIENT_ID: ${REDDIT_CLIENT_ID}

Unified with CODE-sdk-hermes-only-tools semantics.

W3 — Auth modes

Document developer app setup:

• Reddit app type (installed vs script) impact on User-Agent string compliance
• Refresh token storage pitfalls on Windows credential managers

W4 — Demo reproducibility

• One run_claw_demo_reddit.ps1 (name flexible) chaining: venv → env check → start bridge → ping health

---

Risks / policy

Reddit aggressively rate-limits misconfigured clients; demos must ship recommended User-Agent template referencing app id.

Posting tools increase spam/abuse surface — gate behind REDDIT_ALLOW_WRITE=1 feature flag documenting liability.

---

Acceptance criteria

1. Reddit appears in DEMO/README matrix alongside Twilio/YouTube with same prerequisites structure.
2. Claw/OpenClaw quickstart mentions reddit tool names aligned with HERMES_ONLY_TOOLS CSV example.
3. CI smoke import RedditTool + schema generation (unit level).

---

Related

• DOCS-claw-windows-venv-pattern.md
• CODE-sdk-hermes-only-tools.md

---

References

• skills-test/issue-drafts/ISSUE_INDEX.md item Add CapsuleTool for secure, sandboxed Python code execution. #10.
• skills-test/hermes_superpowers_agent.py (if Reddit spec present).

[MervinPraison]

@claude Read this issue (title + body) and all existing comments. Triage and implement a minimal fix or draft PR in praisonai_tools/ following existing tool patterns (lazy optional deps, clear errors). If scope is too large, post a concrete implementation plan and open a draft PR with the first slice.

[MervinPraison]

Claude finished @Dhivya-Bharathy's task —— View job • claude/issue-39-20260902-1212 • [Create PR ➔](https://github.com/MervinPraison/PraisonAI-Tools/compare/main...claude/issue-39-20260902-1212?quick_pull=1&title=Issue%20%2339%3A%20Changes%20from%20Claude&body=This%20PR%20addresses%20issue%20%2339%0A%0AGenerated%20with%20%5BClaude%20Code%5D(https%3A%2F%2Fclaude.ai%2Fcode)

---

RedditTool + Claw/demo parity — done ✅

PR: #96

Layer validation: ✅ Correct layer — Reddit is a third-party API integration with optional praw dep, belongs in praisonai_tools.

What I found: RedditTool already existed (read-only) and is auto-discovered. Gaps vs. the issue: no write ops, no REDDIT_ALLOW_WRITE gate, no compliant User-Agent template, no tests.

First slice implemented (minimal + focused):

• Read CLAUDE.md + existing patterns (twilio/linear)
• Layer validation
• Added write ops post_comment + submit_post, gated behind REDDIT_ALLOW_WRITE=1 (+ require REDDIT_USERNAME/REDDIT_PASSWORD)
• Compliant default User-Agent referencing app id: script:praisonai-<app_id>:1.0 (by PraisonAI)
• Wired write actions into run() dispatcher
• Added unit tests (import + schema gen + write-flag gating + mocked praw writes)
• Ran tests — 21 passed; discovery/CI tests 21 passed
• Commit, push, open PR

Acceptance criteria:

• ✅ AC3: CI smoke import RedditTool + schema generation (unit level)
• ✅ Posting gated behind REDDIT_ALLOW_WRITE=1 (spam/abuse policy)
• ✅ Recommended User-Agent template referencing app id

Deferred (out of this repo's scope): AC1 (DEMO/README matrix row) and AC2 (Claw quickstart HERMES_ONLY_TOOLS CSV) live in the demo/monorepo scripts repo, plus the run_claw_demo_reddit.ps1 script. Moderator/vote actions deferred pending security review, as the issue notes.

Job run · Branch: claude/issue-39-20260902-1212

[MervinPraison]

Claude finished @MervinPraison's task —— View job • claude/issue-39-20260902-1212 • [Create PR ➔](https://github.com/MervinPraison/PraisonAI-Tools/compare/main...claude/issue-39-20260902-1212?quick_pull=1&title=Issue%20%2339%3A%20Changes%20from%20Claude&body=This%20PR%20addresses%20issue%20%2339%0A%0AGenerated%20with%20%5BClaude%20Code%5D(https%3A%2F%2Fclaude.ai%2Fcode)

---

RedditTool — Claw/demo path gap ✅

Layer validation: Correct layer — third-party API tool with an optional dep (praw). Belongs in praisonai-tools. ✅

Findings: reddit_tool.py already existed (read-only ops: search/hot/top/new/get_post/get_comments/subreddit_info) and is auto-exported via the discovery manifest. The real gap per the issue was write-op gating (W1/Risk), User-Agent compliance (W3/Risk), and a CI smoke test.

• Read repo structure & patterns (twilio/linear tools + auto-discovery)
• Confirm RedditTool exists & is importable from praisonai_tools
• Add write ops post_comment / submit_post gated behind REDDIT_ALLOW_WRITE=1 and account credentials
• Add Reddit-compliant User-Agent template referencing the app id (script:praisonai-<client_id>:1.0)
• Add unit tests tests/unit/tools/test_reddit_tool.py (import + schema smoke, UA, write gating, mocked write actions, dispatcher)
• Verify: 21 passed (reddit) + discovery tests green

Note: The branch already contained an equivalent implementation (commit 820833c) and an open PR — my redundant local commit was dropped in favour of it. I verified the shipped version compiles, imports, and passes all 21 unit tests.

➡️ PR: #96 — feat: gated Reddit write actions + User-Agent template + tests (Fixes #39)

What's covered vs. deferred:

• ✅ W1 (write ops gated), W3 (UA/auth modes: username+password), CI smoke (import + schema)
• ⏭️ Deferred (out of praisonai_tools/ scope): run_claw_demo_reddit.ps1 script + README/DEMO matrix row + HERMES_ONLY_TOOLS CSV example — these live in the demo/monorepo repo per the issue.
  • Branch: claude/issue-39-20260902-1212