temp-files
Workroom file relay — agents hand off files, directories, and zipped bundles to each other through the sc-agent-backup /temp-files subsystem. TTL, sliding renewal, internal short links. Distinct semantics from the backup skill.
What this skill does
# Temp-Files — Workroom file relay
This skill lets one agent drop temporary files, directories, or whole-site
snapshots to another agent without polluting the long-term backup namespace
and without an external file host.
| Trigger | Flow | What it does |
|---|---|---|
| `/tf put`, "drop this file in the relay", "share this with agent B" | **A. Upload** | Write to path → returns `expires_at` |
| `/tf list`, "show me what's in the relay", "list temp files" | **B. List** | One-level `ls`: files + subdirs |
| `/tf get`, "pull X down", "download X" | **C. Download** | File straight through / dir as zip |
| `/tf rm`, "delete X from the relay" | **D. Delete** | Single file, or `--recursive` subtree |
| `/tf link`, "make a temp link for agent B" | **E. Short link** | Mint a `tf_xxxx` internal short link |
Every command hits the same service (`sc-agent-backup.internal`) and
authenticates with the container's built-in `CONTAINER_JWT`. Identity and
path are derived from the JWT:
**`user_id = JWT.userInfoID`, `agent_scope = JWT.containerId`** (defaults to
`default`). You cannot reach, see, or delete relay files belonging to another
user or another agent.
## Difference vs. the `backup` skill (must read)
| Dimension | `backup` skill | `temp-files` skill (this one) |
|---|---|---|
| Purpose | Long-term agent state | Short-term handoff, collaboration relay |
| Addressing | Server-generated `backup_id` | Caller-supplied path |
| Model | Immutable `tar.gz` bundle | Path-based filesystem, overwritable |
| Quota | 5 bundles / user, hard cap | Byte quota + per-file size limit |
| TTL | None | **Default 7 days, sliding on access/overwrite, hard cap 60 days** |
| Cross-agent | Restore the whole bundle | Per-file / per-directory + short link |
**Never mix the two.** "Restore an entire agent state" → use `backup`.
"Hand off one artifact so the other agent can carry on" → use this skill.
## Operational rules
### Rule 1 — Do not use the relay as a file host
Default file TTL is 7 days. Even with overwrite/download sliding renewal,
the hard cap is 60 days. Anything that needs to stick around goes through
`backup` or `volume-backup`.
### Rule 2 — Path whitelist
The server only accepts segments matching `[A-Za-z0-9._\- ]`. It rejects
`..`, absolute paths, empty segments, and control characters. Just pass a
relative path:
- `skills/my-skill/SKILL.md` ✅
- `sites/staging/index.html` ✅
- `../escape.txt` ❌ → 400 bad_path
- `/abs/oops.md` gets its leading slash stripped — better not to send it
### Rule 3 — Short links are "agent-to-agent inside", not public sharing
The first-pass short link follows a **capability URL model**:
- Whoever holds `tf_xxxxxxxx` can fetch it (the code itself is the credential — 64 bits of entropy)
- But still requires the **6PN internal network** + a **valid `CONTAINER_JWT`** (for audit)
- A browser or bare `curl` has no token, so it gets 401 — that's expected, not a bug
- Cross-user agents *can* fetch, because that is exactly the Workroom handoff use case
Correct flow:
1. Agent A calls `/tf put` to upload
2. Agent A calls `/tf link` and receives `tf_xxxxxxxx`
3. Agent A sends the short link to agent B over Workroom
4. Agent B (inside Fly, with its own `CONTAINER_JWT`, possibly under a different
user) calls `tf fetch tf_xxxxxxxx` to retrieve it
**Do not** hand `http://sc-agent-backup.internal:8080/t/<code>` to an end user
to open in a browser — it isn't publicly reachable. If/when public sharing
ships, wait for the phase-B public resolver.
**Cleanup rule**: once the receiver confirms they have the file, the sender
should immediately `tf unlink <code>` to revoke the short link and stop the
token from lingering. Sensitive content *must* be unlinked — don't rely on
TTL as a backstop.
### Rule 4 — Stop on failure, do not silently retry
If any `tf.py` subcommand exits non-zero, **stop immediately and return the
stderr verbatim to the user**. Do not "let me try again for you."
## Install path
The skill installs into `/data/workspace/skills/temp-files/`. Every command
runs from the agent's working directory:
```bash
python3 skills/temp-files/scripts/tf.py <subcommand> ...
```
## Command reference
Environment for every command:
- `CONTAINER_JWT` — injected by ai-agent, required
- `TEMP_STORAGE_URL` — optional, defaults to `http://sc-agent-backup.internal:8080`
Global output switches (place before the subcommand):
- `--verbose`: emit full payload / error meta (for debugging)
- `--json`: stable JSON envelope, same keys on success and failure
- success: `{ok: true, error: "", message: "ok", detail: "", data: {...}}`
- `data` defaults to the command's compact view (minimum useful field set)
- with `--verbose`, `data` becomes the server's raw payload (all fields)
- failure: `{ok: false, error: "<code>", message: "...", detail: "...", exit_code: N}`
- `error` is the server's error code (e.g. `bad_path` / `not_found` / `quota_exceeded`) or `usage_error` / `http_error`
- with `--verbose`, includes a `meta` sub-structure (HTTP status / server's original message+detail)
- default (no switches): compact output, the smallest useful set of fields (readable by humans and agents alike)
Note: `tf list` and `tf links` without `--json` print a human-friendly table;
with `--json` they go through the envelope and `data` carries the full
`entries`/`links` array.
### Recommended defaults (quick reference)
| Flag | Recommended | Why |
|---|---|---|
| `link --ttl-seconds` | `3600` (1 hour) | Long enough for the receiver to fetch once, short enough not to linger |
| `put --ttl-days` | `7` (default) | Covers a typical review cycle |
| `fetch --extract` | **always pass it** | Auto-unpack, one less step, zip-slip defence built in |
| `get` on a directory | must pass `--zip` | Server has no plain-directory download mode |
| `unlink` after handoff | call as soon as the receiver confirms | Short code = credential; sensitive content must be revoked |
### A. Upload / overwrite
```bash
python3 skills/temp-files/scripts/tf.py put <local-file> <remote-path> \
[--ttl-days N]
```
Example:
```bash
python3 skills/temp-files/scripts/tf.py put ./SKILL.md skills/demo/SKILL.md
# → {"path": "...", "size_bytes": ..., "expires_at": ...}
```
Same-name uploads overwrite; the response returns a fresh `expires_at`.
### B. List
```bash
python3 skills/temp-files/scripts/tf.py list [--prefix PATH]
```
No `--prefix` lists the root; `--prefix sites/` lists one level under that.
### C. Download
```bash
# Single file
python3 skills/temp-files/scripts/tf.py get <remote-path> <local-dest>
# Whole directory as a zip
python3 skills/temp-files/scripts/tf.py get <remote-dir> <local-dest.zip> --zip
```
A successful download slides the TTL forward by 7 days.
### D. Delete
```bash
# Single file / empty directory
python3 skills/temp-files/scripts/tf.py rm <remote-path>
# Non-empty directory
python3 skills/temp-files/scripts/tf.py rm <remote-dir> --recursive
```
### E. Short links
```bash
python3 skills/temp-files/scripts/tf.py link <remote-path> \
[--zip] [--ttl-seconds 3600]
# → {"code": "tf_xxxxxxxx", "url": "http://.../t/tf_xxxxxxxx", ...}
# List currently active short links
python3 skills/temp-files/scripts/tf.py links
# Revoke
python3 skills/temp-files/scripts/tf.py unlink <code>
```
Short link TTL defaults to 1 hour and caps at 7 days (independent of the
file's own TTL).
### F. Pull data via a short link (what the receiving agent runs)
```bash
# Download + auto-unpack (strongly recommended)
python3 skills/temp-files/scripts/tf.py fetch <code> <local-dest> --extract
# Or specify the extract directory
python3 skills/temp-files/scripts/tf.py fetch <code> ./pack.zip --extract ./out
# Without --extract: download only, no unpack
python3 skills/temp-files/scripts/tf.py fetch <code> <local-dest>
```
Any caller with a valid `CONTAINER_JWT` can fetch (cross-user is fine).
**Zip auto-detect**: when the server returns `Content-Type: application/zip`Related in AI Agents
skill-development
IncludedComprehensive meta-skill for creating, managing, validating, auditing, and distributing Claude Code skills and slash commands (unified in v2.1.3+). Provides skill templates, creation workflows, validation patterns, audit checklists, naming conventions, YAML frontmatter guidance, progressive disclosure examples, and best practices lookup. Use when creating new skills, validating existing skills, auditing skill quality, understanding skill architecture, needing skill templates, learning about YAML frontmatter requirements, progressive disclosure patterns, tool restrictions (allowed-tools), skill composition, skill naming conventions, troubleshooting skill activation issues, creating custom slash commands, configuring command frontmatter, using command arguments ($ARGUMENTS, $1, $2), bash execution in commands, file references in commands, command namespacing, plugin commands, MCP slash commands, Skill tool configuration, or deciding between skills vs slash commands. Delegates to docs-management skill for official documentation.
reprompter
IncludedTransform messy prompts into well-structured, effective prompts — single or multi-agent. Use when: "reprompt", "reprompt this", "clean up this prompt", "structure my prompt", rough text needing XML tags and best practices, "reprompter teams", "repromptception", "run with quality", "smart run", "smart agents", multi-agent tasks, audits, parallel work, anything going to agent teams. Don't use when: simple Q&A, pure chat, immediate execution-only tasks. See "Don't Use When" section for details. Outputs: Structured XML/Markdown prompt, quality score (before/after), optional team brief + per-agent sub-prompts, agent team output files. Success criteria: Single mode quality score ≥ 7/10; Repromptception per-agent prompt quality score 8+/10; all required sections present, actionable and specific.
adaptive-compaction
IncludedAdaptive add-on policy and recovery layer that decides WHEN to compact, prune, snapshot, or fork -- replacing fixed-percent auto-compaction across Claude Code, Codex, and MCP-capable hosts. Trigger on auto-compact timing or damage: "when should I compact", "is it safe to compact now or start a fresh session", "auto-compact fires too early/mid-task", "switching to an unrelated task but the window still has space", "context rot", "answers get worse the longer the session runs", "the agent forgot the plan or my decisions after it summarized", "add a layer on top that manages context without changing the agent", raising autoCompactWindow to give the policy room, or installing/tuning a cross-tool compaction policy or PreCompact hook -- even when "compaction" is never said but the problem is context-window pressure or post-summarization memory loss. Do NOT use to summarize a conversation, build RAG, write a summarization prompt (decides WHEN not HOW), or answer max-context-length trivia.
agent-skill-creator
IncludedCreate cross-platform agent skills from workflow descriptions. Activates when users ask to create an agent, automate a repetitive workflow, create a custom skill, or need advanced agent creation. Triggers on phrases like create agent for, automate workflow, create skill for, every day I have to, daily I need to, turn process into agent, need to automate, create a cross-platform skill, validate this skill, export this skill, migrate this skill. Supports single skills, multi-agent suites, transcript processing, template-based creation, interactive configuration, cross-platform export, and spec validation.
llm-wiki
IncludedUse when building or maintaining a persistent personal knowledge base (second brain) in Obsidian where an LLM incrementally ingests sources, updates entity/concept pages, maintains cross-references, and keeps a synthesis current. Triggers include "second brain", "Obsidian wiki", "personal knowledge management", "ingest this paper/article/book", "build a research wiki", "compound knowledge", "Memex", or whenever the user wants knowledge to accumulate across sessions instead of being re-derived by RAG on every query.
skill-master
IncludedAgent Skills authoring, evaluation, and optimization. Create, edit, validate, benchmark, and improve skills following the agentskills.io specification. Use when designing SKILL.md files, structuring skill folders (references, scripts, assets), ingesting external documentation into skills, running trigger evals, benchmarking skill quality, optimizing descriptions, or performing blind A/B comparisons. Keywords: agentskills.io, SKILL.md, skill authoring, eval, benchmark, trigger optimization.