Shared Pinecone RAG

Use the shared Pinecone RAG index for any agent in this workspace. Use when an agent needs to ingest markdown/text docs into pulse-rag or query semantic cont...

Security Scan

The skill's purpose (shared Pinecone RAG) is plausible, but the instructions reference a hard-coded user path and require a Pinecone API key stored in a .env that the manifest doesn't declare — plus the actual python code run by the scripts (ingest.py / query.py) is not included for review, so the footprint and data flows can't be audited.

Purpose & Capability

The skill claims to provide a shared Pinecone RAG retrieval/ingest layer, which matches the included scripts that run ingest/query. However the SKILL.md requires a PINECONE_API_KEY and a Python venv stored inside a specific user path ('/home/Mike/.openclaw/workspace/rag-pinecone-starter') while the registry metadata lists no required credentials or env vars — that's an incoherence. The absolute path to a specific user's home (Mike) is surprising for a workspace-shared skill and reduces portability.

Instruction Scope

Runtime instructions tell agents to cd into /home/Mike/... and run scripts which activate a .venv and call python ingest.py / query.py. The scripts themselves are tiny wrappers, but the substantive behavior lives in ingest.py and query.py which are not included for review. Those Python programs will likely load the PINECONE_API_KEY from rag-pinecone-starter/.env and may transmit ingested documents to Pinecone — this is expected for a RAG skill but cannot be audited here. The SKILL.md also instructs placing files under that hard-coded docs path, meaning the skill will read host filesystem content in that directory.

Install Mechanism

No install spec — instruction-only with two small shell scripts. This is lower risk from an installation perspective because nothing is downloaded or extracted by the skill itself. However the skill depends on a pre-existing python venv and unseen python scripts.

Credentials

The manifest declares no required environment variables, yet SKILL.md explicitly says 'PINECONE_API_KEY must be set in rag-pinecone-starter/.env' and the scripts activate a .venv (expected). Requiring a secret API key but not declaring it in the skill metadata is an inconsistency. Storing the API key in a repo-local .env is a design choice but increases risk: the key allows write/read access to the shared Pinecone index and could be used to exfiltrate documents if the unseen Python code is malicious or misconfigured.

Persistence & Privilege

The skill is not marked always:true and does not request elevated persistent presence. Autonomous invocation is allowed (default). Because the skill has access to a network-capable pinecone API key (per SKILL.md) and reads local docs, autonomous invocation combined with an undisclosed credential increases blast radius — verify credentials and code before letting the agent call this skill without supervision.

What to consider before installing

This skill is plausible for sharing a Pinecone-based RAG, but there are several red flags to check before installing or invoking it: - The manifest does not declare PINECONE_API_KEY or any required env vars, but SKILL.md says the key must be set in rag-pinecone-starter/.env. Confirm where the key lives and whether it should be explicitly declared in the skill manifest. - The instructions and scripts use a hard-coded absolute path (/home/Mike/.openclaw/workspace/rag-pinecone-starter). Ensure the path is correct for your environment and not pointing to another user's home. Prefer a workspace-relative path. - The actual behavior is inside ingest.py and query.py, which are not included here. Do NOT run the scripts until you inspect those Python files to confirm they only contact Pinecone (or other expected endpoints), handle secrets safely, and do not exfiltrate data to unknown endpoints. - Treat the Pinecone API key as sensitive. If you proceed, limit the key's permissions where possible and rotate it if you later remove the skill. - Consider running the ingest process in an isolated environment (container or dedicated service account) and test with non-sensitive data first. If you can provide ingest.py and query.py (or modify the skill to declare the required env var and use workspace-relative paths), the assessment could be upgraded to benign if their code matches the stated purpose.

Use the central RAG project at: /home/Mike/.openclaw/workspace/rag-pinecone-starter

When combined with hybrid-db-health, position this as a Persistent Memory skill stack:

• shared-pinecone-rag = retrieval + ingest layer
• hybrid-db-health = reliability/health guardrail layer

Query (all agents)

bash scripts/query-shared-rag.sh "your question"

Ingest docs (all agents)

1. Put .md/.txt files in: /home/Mike/.openclaw/workspace/rag-pinecone-starter/docs/
2. Run:

bash scripts/ingest-shared-rag.sh

Requirements

• PINECONE_API_KEY must be set in rag-pinecone-starter/.env
• Python venv exists at rag-pinecone-starter/.venv

Notes

• Index name defaults to pulse-rag.
• Retrieval reads from namespace default.
• This skill is shared; do not duplicate per-agent RAG stacks unless explicitly requested.