Skip to content
Security
Command

/plan-evidence-transfer

Plan a human-approved secure evidence transfer without collecting or transmitting evidence

From plugin
trust-center
433146 skills146 commands1 MCP
> /plugin marketplace add GRCEngClub/claude-grc-engineering

Command definition

plan-evidence-transfer.md
description: Plan a human-approved secure evidence transfer without collecting or transmitting evidence

Plan evidence transfer

Prepare a **plan**, not a package or transmission. This is an initial manual workflow toward secure evidence packaging, not an encryption or delivery implementation.

Arguments

  • `$ARGUMENTS` - Intended recipient and purpose, classification, approximate size, and approved delivery constraints. Use generic descriptions; do not paste evidence, secrets, credentials, private keys, access tokens, or personal data into the conversation.

Instructions

1. Ask for missing information: authorized recipient and independent verification contact, permitted classification and approved transfer method, retention/deletion requirements, and whether signing is required. Do not assume that a portal or email is approved. 2. Produce an inventory template containing only artifact identifiers/descriptions, owner, classification, intended recipient, and retention date. Explain that names, paths, metadata, hashes, and even the inventory can be sensitive. The evidence owner must select the minimum necessary artifacts and review them for secrets and third-party data before inclusion; this workflow does not scan or redact them. 3. Recommend use of an organization-approved, access-controlled workstation and private encrypted staging location. Avoid shared temporary directories, repositories, chat transcripts and logs. Review file permissions, symlinks, unexpected files and archive paths before packaging; do not follow links into unrelated locations. A normal archive is not encrypted. Archive metadata may reveal sensitive information. 4. If asymmetric encryption is approved, have the recipient provide a supported public key and verify its **full fingerprint out of band** with a known contact before use. Determine approved signing keys and fingerprint verification separately. Use the organization's supported GnuPG/PGP implementation and documented key-management procedures; never invent cryptography or request a private key, passphrase or recovery code. A public key in an unverified email is not sufficient. 1Password CLI is not treated as an encryption format. 5. Within the approved secure workspace, the operator can create a manifest of selected files and their SHA-256 digests, then package the reviewed files and manifest using approved tooling. Have an authorized operator sign the manifest or package when required and encrypt the archive to the independently verified recipient key(s). Confirm signer and recipient fingerprints and verify the output with the organization's documented tools before release. Hashes alone do not authenticate the sender; the recipient must validate the signature using a separately verified signing key and check the file digests after safe extraction. 6. Obtain explicit human approval of the recipient, contents and delivery channel before any transfer. A human uses the approved portal or transfer tool; do not upload, email, publish links, change permissions or contact recipients automatically. Record the transfer reference, time, recipient identity, key fingerprints, package digest and approval in the approved restricted audit system, without credentials, evidence content or secret-bearing links. Confirm receipt and signature/checksum verification through an agreed channel. 7. Apply the organization's retention and cleanup policy to staging files and keys. Do not promise secure deletion: backups, snapshots, SSDs and synchronized folders can retain copies. Record exceptions and failed verification; do not release a failed package.

Local packaging (optional)

The operator may use the experimental local CLI below after completing the planning and key verification steps above. Run it directly on a trusted, access-controlled machine, not through a chat containing evidence or credentials. Import the recipient's public key into the operator's GnuPG keyring first and independently verify the full **primary key** fingerprint. The CLI deliberately uses GnuPG's `--trust-model always` only after requiring that exact fingerprint; it cannot verify the recipient's identity for you.

python3 plugins/grc-engineer/scripts/package-evidence.py /private/reviewed-evidence /private/output/evidence.zip.gpg FULL_PRIMARY_KEY_FINGERPRINT

The source must be a stable, trusted directory, not writable by other users or processes during packaging. The output must not already exist or be within the source. The command stores regular files in a ZIP, includes a SHA-256 JSON manifest inside it, and encrypts to one GnuPG recipient. It refuses symlinks and special files; empty directories are omitted. ZIP timestamps use local time with two-second precision: dates before 1980 are clamped to 1980-01-01 00:00:00 and dates after 2107 to 2107-12-31 23:59:58 rather than preserved. Permission bits are stored where supported, but other metadata is not. File contents and the archive are held in memory, so use only for reasonably sized evidence. It creates no plaintext archive on disk, but memory, swap, crash dumps and backups can expose plaintext. The output digest is printed for the operator's restricted audit record. File names, metadata, hashes and manifest can be sensitive.

Before encryption the packager runs the offline verifier against its in-memory ZIP. It refuses unsafe archive paths and archives exceeding the verifier's limits, including the manifest; no output is created on verification failure. The check runs only after the ZIP is constructed in memory, so these limits do not prevent excessive memory use during packaging. This adds an internal consistency check, not sender authentication. Use a trusted, stable output parent directory, not writable by untrusted users: output creation is exclusive but publication is not atomic and an observer may see a partial file during writing.

This initial CLI does **not** sign the manifest or authenticate the sender, verify re

Read more
@grcengclub
Ships withtrust-center

Open-source GRC Engineering resource for Claude. claude-grc-engineering turns technical evidence from cloud, SaaS, code, and security tools into framework-aligned findings, gap reports, remediation guidance, evidence packages, and OSCAL workflows.

433105 skills1 MCP
Get the whole plugin

Stars on GitHub

Momentum

stars in total
433
Star history of GRCEngClub/claude-grc-engineering on GitHub, a chart by star-history.com

Other commands on trust-center.

research

research

Command

Start or resume an academic research project — idea through literature, methodology, writing, feedback, and publishing

collect

collect

Command

Query AWS for compliance-relevant configuration across IAM, S3, CloudTrail, EBS, and emit findings conforming to the v1 contract.

setup

setup

Command

Install the frdocx-to-froscal-ssp Python pipeline and verify its dependencies. Idempotent.

retrieve

retrieve

Command

Retrieve a single AWS Secrets Manager secret value to stdout or a 0600-permission file. Opt-in retrieval mode — never writes to…

More commands from other plugins.

deslop-reborn

deslop-reborn

Command

One iteration of the IronClaw Reborn de-slop loop — take ONE Reborn crate, fan out parallel review sub-agents (thermo-nuclear…

@nearai@nearaiInstall Command
audit

audit

Command

Audit a file, directory, or whole repo for insecure default configuration: fallback secrets, default credentials, fail-open…

@trailofbits@trailofbitsInstall Command
arsenal

arsenal

Command

Show which external bug-bounty tools are installed on this machine and print install hints for the missing ones. Curated from…

@awarexone@awarexoneInstall Command