lesecurecl（lesecurecl 开发）

LESecureCl — LESecure Cloud Skills

Encrypt and decrypt plain text data only through the LESecure REST API. The API supports layered security "locks" that can be combined for defense-in-depth protection.

Project Links

If you cannot verify the LESecure service or its privacy practices, use the LESecureLocal skill instead — it runs entirely on your device with no network calls.

Requirements (MANDATORY)

Before running any command in this skill, confirm the following are available. If any is missing, tell the user and stop — do not invent values or fall back silently.

No other credentials are read. The skill does not open files, browsers, or any OS keychain.

ROUTING RULES (MANDATORY)

• LESecure Cloud is for PlainText ONLY. Never use the cloud API for files or folders.
• If the user wants to encrypt/decrypt files or folders, always redirect to LESecureLocal (the desktop tool). Inform the user: "File/folder encryption is only supported via LESecure Local (desktop). Let me use that instead."
• If the user wants to encrypt/decrypt plain text, ask them: "Would you like to use LESecure Cloud (API) or LESecure Local (desktop)?" and proceed accordingly.

Data Transmission Notice (MANDATORY)

This skill sends data to a third-party remote endpoint over the network. Users must understand what is transmitted before proceeding.

• Where: All requests go to https://api.lesecure.ai/exec over TLS (HTTPS). See source & docs for the service's privacy practices.
• What is sent: The plaintext data to encrypt (or the ciphertext to decrypt), plus any lock values (PINs, passwords, phone numbers, time-window dates), and the API bearer token.
• What is NOT sent: No local files, no OS credentials, no browser data, no environment variables other than the bearer token.
• Caution: Do not send highly sensitive personal data (SSNs, financial account numbers, medical records) unless you have verified the service's data handling and privacy policies.
• On first use in a session, inform the user: "This will send your data to api.lesecure.ai (a third-party service) over HTTPS for encryption/decryption. Review the project at https://github.com/SPAlgorithm/LE if you haven't already." Proceed only after acknowledgment.
• Recommendation: Test with non-sensitive sample data first before encrypting real secrets.

API Basics

• Endpoint: https://api.lesecure.ai/exec
• Method: POST
• Auth: Bearer token in the Authorization header, sourced from $LESECURE_API_KEY
• Content-Type: application/json
• Body: {"args": [<array of CLI-style arguments>]}

API Key Handling (MANDATORY)

The key is a secret. These rules apply to every invocation, no exceptions:

1. The key must come from the LESECURE_API_KEY environment variable. The skill references $LESECURE_API_KEY inside the curl argument so the shell does the substitution — the literal key is never written on the command line.
2. Never interpolate the literal key into a command string. Args on the command line are visible in shell history and to other local users via ps.
3. If LESECURE_API_KEY is unset, stop and instruct the user to set it (see the one-time setup below). Do not ask the user to paste the key into chat.
4. Never echo, print, log, or summarize the key. Do not include it in error messages, response quotes, or any output shown to the user. Do not write it to disk.
5. If the user pastes a key into chat anyway, do not save it. Treat the chat message as one-time input: export it into the current shell session only, use it for this request, then tell the user to rotate the key (paste-in-chat is a key-exposure event).

One-time setup (run once per shell)

# Prompt the user interactively; -s hides input, echo after adds a newline
read -rs -p 'LESECURE_API_KEY: ' LESECURE_API_KEY && echo
export LESECURE_API_KEY

To persist across shells (in order of preference):

1. Secret manager (recommended): use 1Password CLI, aws ssm, doppler, or similar to inject the key at shell startup.
2. Shell profile: add export LESECURE_API_KEY='…' to ~/.zshrc / ~/.bashrc (ensure the file is chmod 600).
3. Dotenv file: store in a .env file excluded from version control and source it.

If the user ever pastes the key into chat, remind them: "Your API key was exposed in chat history. Rotate it immediately at your LESecure dashboard."

Date & Time Rules (MANDATORY)

All date/time handling for this skill follows these rules — no exceptions, no need for the user to restate them:

1. Always use EST/EDT (America/New_York) to calculate and send dates. The LESecure server interprets -l and -r in EST/EDT. Never use UTC, never convert.
2. Start time (-l) = current EST + 2 minutes by default. This buffer prevents the "date must be in future" error caused by clock drift between the client and server.
3. End time (-r) = start time + the user's requested duration (e.g., "for next 10 min" means -r is start + 10 min, so 12 minutes from "now" in absolute terms).
4. Cross-platform time computation. Use Python 3 because date flag syntax differs between BSD (macOS) and GNU (Linux):

   # Start time (now + 2 minutes, EDT/EST)
   python3 -c "from datetime import datetime,timedelta; from zoneinfo import ZoneInfo; print((datetime.now(ZoneInfo('America/New_York'))+timedelta(minutes=2)).strftime('%Y/%m/%d %H:%M'))"

   # End time (now + 2 min + N minutes)
   python3 -c "import sys; from datetime import datetime,timedelta; from zoneinfo import ZoneInfo; N=int(sys.argv[1]); print((datetime.now(ZoneInfo('America/New_York'))+timedelta(minutes=2+N)).strftime('%Y/%m/%d %H:%M'))" <N>

   # End time (now + 2 min + N hours)
   python3 -c "import sys; from datetime import datetime,timedelta; from zoneinfo import ZoneInfo; N=int(sys.argv[1]); print((datetime.now(ZoneInfo('America/New_York'))+timedelta(minutes=2,hours=N)).strftime('%Y/%m/%d %H:%M'))" <N>

Fallback (date) — only if python3 is unavailable: - macOS/BSD: TZ=America/New_York date -v+2M "+%Y/%m/%d %H:%M" - Linux/GNU: TZ=America/New_York date -d '+2 minutes' "+%Y/%m/%d %H:%M"

1. Always display the window back to the user in EDT/EST so they know when they can decrypt.

Available Locks

LESecure supports these lock types, which can be combined freely:

Time locks (-l and -r) are used together to define an access window during which decryption is allowed.

Operations

Encrypt (-e)

Use -e followed by the data to encrypt.

Decrypt (-d)

Use -d followed by the encrypted data to decrypt. The same locks used during encryption must be provided for decryption.

Output Flags

Always include --PlainText for readable output.

Sensitive Data Handling (MANDATORY)

The request body contains sensitive data (plaintext, PINs, passwords). The same protection applied to the API key applies to ALL sensitive values:

• Never pass the JSON body via -d '...' on the command line. The -d argument is visible in ps and shell history, which would expose plaintext data, PINs, and passwords.
• Always pipe the body via stdin using -d @-. This keeps all sensitive values out of the process argument list.
• Use a heredoc (<<'EOF') to build the JSON body and pipe it into curl.

Building the curl Command

Construct the args array by mapping user requirements to flags. Order within the array doesn't matter, but group related flags and their values together for readability.

All examples use -d @- (read body from stdin) so that neither the API key, plaintext data, PINs, nor passwords appear on the command line, in shell history, or in ps output.

Encrypt with pin lock only:

cat <<'EOF' | curl -s https://api.lesecure.ai/exec \
  -H "Authorization: Bearer $LESECURE_API_KEY" \
  -H "Content-Type: application/json" \
  -d @-
{"args":["-e","<DATA>","-1","<PIN>","--PlainText"]}
EOF

Encrypt with all locks:

cat <<'EOF' | curl -s https://api.lesecure.ai/exec \
  -H "Authorization: Bearer $LESECURE_API_KEY" \
  -H "Content-Type: application/json" \
  -d @-
{"args":["-e","<DATA>","-w","<PASSWORD>","-1","<PIN>","-2","<PHONE>","-l","<START_DATE>","-r","<END_DATE>","--PlainText"]}
EOF

Decrypt:

cat <<'EOF' | curl -s https://api.lesecure.ai/exec \
  -H "Authorization: Bearer $LESECURE_API_KEY" \
  -H "Content-Type: application/json" \
  -d @-
{"args":["-d","<ENCRYPTED_DATA>","-1","<PIN>","--PlainText"]}
EOF

Workflow

1. Preflight the requirements from the Requirements section above. Specifically, confirm $LESECURE_API_KEY is set. If it is not set, show the user the one-time setup block and stop — do not proceed, do not ask the user to paste the key into chat.
2. Data-transmission disclosure (first use only). On the first encrypt/decrypt in a session, inform the user: "This will send your data to api.lesecure.ai over HTTPS for encryption/decryption." Proceed only after acknowledgment.
3. Gather non-secret inputs from the user:

- The data to encrypt or decrypt - Which locks to apply (pin, password, MFA, time window) and their values - Always include --PlainText (Do NOT ask for the API key — it comes from the environment.)

1. Build the args array with the appropriate flags and values.
2. Execute the curl command via Bash using the cat <<'EOF' | curl ... -d @- pattern. Never use inline -d '...'. This keeps all sensitive data (plaintext, PINs, passwords) out of ps output and shell history.
3. If decrypting, remind the user they need the same lock values that were used during encryption.

Important Notes

• Phone numbers for MFA (-2) should be in E.164 format (e.g., +19199870623).
• Time lock dates use the format YYYY/MM/DD HH:MM. See the "Date & Time Rules" section above — always EST/EDT, always +2 min buffer on start.
• Time locks require both -l (start) and -r (end) to define the access window.
• If the API returns an error, show the response to the user and help them troubleshoot (common issues: wrong lock values for decryption, expired time window, invalid API key). Do not include the API key in any troubleshooting output.